AI Search Optimisation: 100 Questions on GEO, AEO, llms.txt, AI Overviews and Getting Recommended by ChatGPT, Answered

By Yamm Labs · Updated September 2026 · 35 min read

Short answer: This page answers 100 questions that marketers, founders and SEO teams type into Google, ChatGPT, Perplexity and Gemini about AI search optimisation: what GEO and AEO are, how AI engines choose sources, how to get a business recommended by ChatGPT, what llms.txt and agents.md do and do not do, how robots.txt should treat AI crawlers, which content formats get cited, how to measure AI visibility, and how all of it applies to banks, insurers and fund houses. Each answer is self-contained and cites its source where a third-party fact is used. Yamm Labs is a design-led brand agency for fintech and BFSI companies in Gurugram, India, founded in 2017. It runs AI visibility (GEO/AEO) programmes for BFSI brands.
On this page
  1. What is AI search optimisation, and what else is it called?
  2. How do AI engines decide which sources to retrieve and cite?
  3. How do you get ChatGPT and other AI engines to recommend your business?
  4. What are llms.txt and agents.md, and do they actually work?
  5. How should robots.txt handle AI crawlers?
  6. How do structured data and entity consistency affect AI search?
  7. Which content formats get cited by AI engines?
  8. Why do reviews, listicles, Reddit and other third-party sources matter so much?
  9. How do you measure AI visibility, and which tools help?
  10. How does AI search optimisation work for BFSI and regulated brands?
  11. Sources

How to use this page: each section is ten questions on one theme. This hub is a companion to our shorter GEO and AEO FAQ (60 questions), which covers per-engine source selection and BFSI compliance basics; the questions here are new and do not repeat it. Where an engine’s behaviour is described, it is a working model built from the publishers’ own documentation and our monthly re-queries, not a published specification.

What is AI search optimisation, and what else is it called?

What is AI search optimisation?

AI search optimisation is the practice of making a business, and the pages that describe it, easy for AI answer engines to find, understand, trust and cite. It covers ChatGPT, Perplexity, Gemini, Google AI Overviews and AI Mode, Copilot and Claude. Traditional SEO aims at a ranking position; AI search optimisation aims at being named or cited inside a generated answer. The work is content structure, entity consistency, crawler access and third-party corroboration, in that order. Yamm Labs runs this for BFSI brands from Gurugram.

What is AI search optimisation also known as?

AI search optimisation is also called generative engine optimisation (GEO), answer engine optimisation (AEO), LLM optimisation (LLMO) and sometimes AI SEO or AI visibility. The names overlap heavily. GEO emphasises engines that generate a written answer; AEO emphasises direct answers to a question, including featured snippets and voice; LLMO emphasises the language model itself. Buyers use the terms interchangeably, so pick one for your own documents and treat the rest as synonyms when you brief an agency.

What is the acronym for AI search optimisation?

There is no single accepted acronym for AI search optimisation. GEO (generative engine optimisation) is the most common in agency and research writing, AEO (answer engine optimisation) is common in marketing tools, and AIO or AISO appear occasionally. If you are writing a job description or a brief, say “AI search optimisation (GEO/AEO)” once and then use whichever you prefer. The acronym matters less than the deliverables: citations, mentions and correct facts in AI answers.

When did generative engine optimisation and answer engine optimisation start?

Answer engine optimisation predates generative AI: the term was used for featured snippets and voice assistants in the late 2010s. Generative engine optimisation became common in late 2023, after a research paper titled “GEO: Generative Engine Optimization” (arXiv, November 2023) tested how content changes affect visibility in generated answers, and after ChatGPT gained web browsing. Google’s AI Overviews launched broadly in 2024 and AI Mode followed, which is when most Indian marketing teams first budgeted for the work.

Is generative engine optimisation real, or is it rebranded SEO?

Generative engine optimisation is real in the sense that AI engines demonstrably choose some pages and brands over others, and that structure, clarity and third-party corroboration change which ones get chosen. It is also true that most of the work is good SEO done properly: crawlable pages, clear entities, honest content, earned mentions. What is new is the unit of success. SEO measured a ranking; GEO measures whether your brand appears, and how it is described, in a generated answer.

What is AI search, as distinct from a search engine?

AI search is any product that answers a question in generated prose rather than with a list of links. The system retrieves candidate pages, reads passages, and writes an answer with citations. Google AI Overviews and AI Mode, ChatGPT search, Perplexity, Copilot and Claude with web search all work this way. A classic search engine ranks documents; AI search selects passages and synthesises. That is why one clear paragraph can outperform an entire long page.

How does an AI search algorithm work, in plain terms?

An AI search algorithm typically runs in three stages. First, the engine rewrites your question into several sub-queries and runs them against a web index, its own or a partner’s. Second, it pulls the most relevant passages from the top results and ranks them for usefulness and trust. Third, a language model writes the answer from those passages and attaches citations. Optimisation targets stage two: being in the index, being retrievable for the sub-queries, and having quotable passages. This is a working model, not a published specification.

What is generative engine marketing?

Generative engine marketing is a broader label for using AI answer engines as a marketing channel: optimising to be cited (GEO), monitoring how the brand is described, correcting wrong facts, and shaping the third-party sources the engines lean on. It sits between SEO, PR and brand management. For a regulated brand, it also includes making sure every statement an engine might quote is compliant, because you cannot attach a disclaimer to a paraphrase.

Is SEO still relevant when AI answers the question?

Yes. AI engines retrieve from web indexes, so a page that cannot be crawled, indexed or ranked cannot be cited. Google states that there are no additional requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible for a snippet. What changes is the payoff: fewer clicks per query, but a citation carries the brand into the answer itself. Keep the SEO foundations and add answer-first structure, entity consistency and third-party mentions on top.

What is the difference between AI search optimisation and using AI tools for SEO?

AI search optimisation is about where your brand appears: inside answers generated by ChatGPT, Gemini, Perplexity and Google’s AI features. Using AI tools for SEO is about how you work: drafting briefs, clustering keywords or generating meta descriptions with a model. They are often confused because both contain the word AI. A team can use no AI tools and still do AI search optimisation well, and a team can use many AI tools and remain invisible in AI answers.

How do AI engines decide which sources to retrieve and cite?

Does ChatGPT use its own index or a partner search engine?

ChatGPT search uses OpenAI’s own crawler, OAI-SearchBot, which OpenAI describes as used to surface websites in search results in ChatGPT’s search features. OpenAI has also drawn on third-party search results at various times, so the practical assumption is a blend. What matters for a site owner is simple: allow OAI-SearchBot in robots.txt, keep pages indexable, and check server logs for its visits. If the bot never fetches your pages, ChatGPT search cannot cite them.

What is retrieval-augmented generation, and why does it matter for visibility?

Retrieval-augmented generation (RAG) is the pattern where a language model is given fresh documents retrieved for the question and writes its answer from them rather than from memory alone. Every AI search product uses some form of it. It matters because the model can only cite what was retrieved, and retrieval is a search problem: relevance, freshness, crawlability and trust. Optimising for RAG means being retrievable for the sub-questions and quotable once fetched.

What is query fan-out, and how does it affect which pages get cited?

Query fan-out is the step where an AI engine expands one question into several narrower searches, for example “fintech branding agency India” into cost, examples, regulation and reviews. In our working model, all the major engines do some version of this. The consequence is that a page can be cited for a sub-question it answers well even if it would never rank for the original phrase. Pages with clearly labelled sections, each answering one sub-question, are retrieved more often.

Do AI engines read whole pages or individual passages?

In practice, passages. An AI engine fetches a page, splits it into chunks, and scores each chunk against the sub-queries. The winning chunk is what gets quoted or paraphrased, with the page as the citation. That is why a 150-word section with a question heading and a direct first sentence is more useful than a 3,000-word essay with the answer buried in paragraph nine. Write each section so it makes sense when lifted out on its own.

Why does ChatGPT recommend a business it was never told about?

Because it was told, indirectly. When ChatGPT names a business, the name usually came from retrieved web pages: a listicle, a directory, a review site, a news mention, the company’s own site or a Reddit thread. Sometimes it comes from training data, which is older and less controllable. If ChatGPT is recommending your business, ask it which sources it used, then protect and strengthen those sources rather than assuming the recommendation is permanent.

Does Google use AI in normal search results, not just AI Overviews?

Yes. Google has used machine-learned systems in ranking for years, and AI Overviews and AI Mode add a generated layer on top. For site owners the practical point is that the same index feeds both. Google’s documentation says there are no additional requirements to appear in AI Overviews or AI Mode, and that you do not need to create new machine-readable files, AI text files or markup. If your page is indexed and eligible for a snippet, it is eligible for the AI features.

How do you get a page to appear in Google AI Overviews?

Start with what Google publishes: the page must be indexed, eligible for a snippet, and follow normal Search policies; no special schema is required. Then do what AI Overviews visibly reward: a direct answer near the top, headings phrased as the query, supporting detail underneath, clear dates and a named organisation. Check the queries where an AI Overview already appears and see which passages are cited. Then write the better version of that passage.

Does a featured snippet help a page get into AI Overviews?

Often, but not automatically. Featured snippets and AI Overviews both favour short, direct passages that answer a specific question, so a page that wins snippets is usually well structured for AI features too. But AI Overviews synthesise from several sources and may cite pages that never held a snippet. Treat snippet-style writing as the foundation, then add depth: comparison tables, definitions and worked examples that give the engine more quotable material per page.

Do AI engines prefer fresh content?

For time-sensitive queries, yes; for definitions and evergreen questions, less so. Engines that search live (Perplexity, ChatGPT search, AI Mode) can weight recency heavily when a question implies it, such as “best in 2026” or anything about rules and prices. A visible last-updated date and a changelog help the engine judge freshness. The practical habit is a quarterly review of your key pages, updating facts and the date, rather than publishing new pages for their own sake.

Do AI engines use the same signals as Google rankings?

Partly. Engines that rely on a web index inherit its ranking to choose candidates, so authority, relevance and crawlability still matter. After that, the selection is different: the engine wants passages that answer the sub-question directly, agree with other sources, and come from an entity it can identify. A page can rank fifth and be cited first because it has the clearest paragraph. Consistency across sources is a signal AI engines weigh more heavily than classic ranking does.

How do you get ChatGPT and other AI engines to recommend your business?

How can I get ChatGPT to recommend my business?

To get ChatGPT to recommend your business, make it retrievable and corroborated. Publish a clear entity page that states what you do, for whom, where, and since when. Allow OAI-SearchBot in robots.txt. Get the same description onto third-party pages ChatGPT already cites for your category: directories, listicles, review sites, trade press. Then ask ChatGPT the prompts your buyers use and note which sources it names. Fix or strengthen those sources first. Repeat monthly.

ChatGPT is recommending my business. How did that happen?

ChatGPT is recommending your business because retrieved pages, or its training data, associated your name with the query. Ask it to list its sources in the same chat; it will usually show the citations. Typical answers are a directory profile, an industry listicle, a Reddit thread or your own service page. Save those URLs. Your job now is to keep them accurate and add two or three more corroborating sources, because a single listicle can be edited or removed.

How long does it take for AI engines to know about a new business?

For engines that search live, days to weeks after the pages are indexed and third-party mentions exist. Perplexity and ChatGPT search can cite a page shortly after their crawlers fetch it. Google AI Overviews depend on Google’s index, so normal indexing timelines apply. Answers drawn from training data lag by months or more, and you cannot accelerate them. Plan for the live-search engines first and treat training-data answers as a slow trailing indicator.

Can I pay to be recommended by ChatGPT, Perplexity or Gemini?

No engine currently sells placement inside its organic generated answer, and no agency can guarantee it. What you can buy is adjacent: sponsored listings on directories that engines cite, paid placements in publisher listicles, and advertising on the platforms. Those may help indirectly by creating corroborating sources, but they must be disclosed. Be wary of any vendor promising a guaranteed mention. See the note on why no agency can guarantee placement in our GEO and AEO FAQ.

Does my business need a Wikipedia page to be recommended by AI?

No. Most businesses recommended by AI engines have no Wikipedia page. Wikipedia and Wikidata help engines disambiguate well-known entities, and a Wikidata entry is worth having if you meet its notability rules, but small and mid-sized firms are recognised through their own site, LinkedIn company page, Google Business Profile, directories and press. Do not attempt to create a promotional Wikipedia article; it will be deleted and may damage trust.

Which prompts should we test to see if AI engines recommend us?

Test the prompts buyers actually type, not your brand name. Build a set of 20 to 40: category prompts (“fintech branding agency in India”), problem prompts (“how to rebrand a bank without losing trust”), comparison prompts (“X vs Y for insurance marketing”) and location prompts (“branding agency in Gurugram for BFSI”). Run each in ChatGPT, Perplexity, Gemini and AI Mode, record whether you are Present, Partial or Absent and which sources are cited. That set becomes your baseline.

What makes a business look legitimate to an AI engine?

An AI engine judges legitimacy by consistency and corroboration. The same name, address, founding year, services and phone number on your site, LinkedIn, Google Business Profile and directories. A physical address and a named location. Third-party mentions that repeat the same facts. Clear pages about who you are and who your clients are. Contradictions, missing contact details, and claims that appear nowhere else make an engine hedge or skip you.

Should we name competitors on our own website?

Yes, carefully. Comparison and alternatives pages are among the most cited formats because engines answering “X vs Y” or “best agencies for Z” need a page that already frames the comparison. Name real competitors, describe honestly when they are the better choice, and keep facts attributed. Avoid disparagement. Our best fintech branding agencies in India post is an example of a first-party listicle that names other agencies alongside Yamm Labs.

Does my location matter for AI recommendations?

Yes. Prompts often carry a place (“in Gurugram”, “in India”, “in the UAE”), and engines prefer entities whose location is explicit and consistent. State your city on your home page, About page, contact page and schema, and keep it identical across LinkedIn and Google Business Profile. Yamm Labs, for example, states Gurugram, Haryana on every entity page. If you serve other markets, say so in a sentence rather than listing cities as offices you do not have.

What is a simple AI search optimisation checklist for a small business?

Seven items. One entity page with name, location, founding year, services and clients. Consistent details on LinkedIn, Google Business Profile and two directories. A robots.txt that allows OAI-SearchBot, PerplexityBot and Claude-SearchBot at minimum. Ten service and FAQ pages written answer-first with question headings. Three third-party mentions that repeat your description. A 20-prompt baseline across four engines. A monthly re-query with a fix list. That is enough to start and to know whether more is worth it.

What are llms.txt and agents.md, and do they actually work?

What does an llms.txt file actually do?

An llms.txt file is a Markdown file at /llms.txt that gives language models a concise map of a site: a title, a short summary and lists of links with notes. The proposal, published by Jeremy Howard in September 2024, says agents are best served by concise, expert-level information gathered in a single accessible location. It does not control crawling, does not affect rankings and does not push content into answers; it only makes the site easier to navigate for an agent that chooses to read it.

Does llms.txt actually work? Which engines use it?

Honestly, the evidence is thin. No major search engine has stated that llms.txt affects retrieval or citation, and Google’s documentation says explicitly that you do not need to create new machine-readable files or AI text files to appear in AI Overviews or AI Mode. Where llms.txt is useful today is for coding assistants and agents that fetch documentation directly. Publish it because it costs an hour, but do not expect it to move visibility on its own.

Is llms.txt a standard?

No. llms.txt is a proposal, not a standard adopted by a standards body or by the major engines. The specification lives at llmstxt.org, first published in September 2024 with a second version in August 2026. It is widely implemented by documentation sites and developer tools, which gives it de facto momentum, but adoption by a site does not obligate any engine to read it. Treat it like a sitemap for agents: helpful, optional, unenforced.

What is llms-full.txt, and how is it different from llms.txt?

llms-full.txt is a companion convention: where llms.txt is a short index of links and summaries, llms-full.txt contains the full text of the key pages concatenated into one Markdown file so an agent can read everything in a single fetch. It is popular for software documentation. For a marketing site a full file is rarely needed; a clean llms.txt pointing to your entity page, services and FAQs does the job. Keep both in sync if you publish both.

How do you create an llms.txt file?

Write a Markdown file. The specification requires an H1 with the site or project name, then an optional blockquote summary, then optional H2 sections each containing a list of links in the form [name](url) followed by optional notes. Put the most important pages first: your entity page, services, FAQs, pricing if public. Save it as UTF-8 plain text and serve it at your domain’s /llms.txt path. A generator tool is unnecessary; a text editor is enough.

How do you add llms.txt to WordPress or Shopify?

On WordPress, upload the file to the web root by FTP or your host’s file manager, or use a plugin that serves a virtual llms.txt; several SEO plugins now include the option. On Shopify you cannot write to the root directly, so use an app that serves the file, or a redirect from /llms.txt to a hosted text file. In both cases, fetch the URL in a browser afterwards to confirm it returns plain text, not a themed 404 page.

What is the difference between llms.txt and robots.txt?

robots.txt is a decades-old, widely honoured file that tells crawlers which paths they may fetch; it can allow or block GPTBot, PerplexityBot or ClaudeBot. llms.txt is a new, optional file that tells a model what the site is about and where the important pages are. One is access control, the other is a guide. If robots.txt blocks a crawler, that crawler will never read your llms.txt. Get robots.txt right first.

Is llms.txt a replacement for schema markup or meta keywords?

No. Schema markup (JSON-LD) is read by Google and others to understand entities and page types, and it is a mature, documented signal. Meta keywords are ignored by major engines and have been for years. llms.txt is neither; it is a navigation file for agents. If you have to choose, keep your Organization, Article and FAQPage schema accurate and ignore meta keywords. Add llms.txt as a low-cost extra, not as a substitute.

How do you check whether an llms.txt file is valid or working?

Fetch your domain’s /llms.txt directly and confirm three things: it returns HTTP 200, the content type is plain text or Markdown, and it starts with a single H1. Then check server logs for fetches of that path by AI user agents. Free online validators exist but they only check format. The only real test of working is whether an agent given your domain finds the right pages faster, which you can test by asking a browsing model to summarise your site.

What is agents.md, and is it related to llms.txt?

AGENTS.md is an open format described as a README for coding agents: a file in a software repository that tells tools such as Codex, Jules, Cursor and Copilot how to build, test and style the code. It is stewarded by the Agentic AI Foundation under the Linux Foundation. It has nothing to do with search engines, AI Overviews or marketing visibility, and it is unrelated to llms.txt beyond both being plain-text files for AI systems. A marketing site needs only llms.txt, if that.

How should robots.txt handle AI crawlers?

Which AI crawlers should a business know about?

The ones that matter for visibility, by publisher: OpenAI’s GPTBot (training), OAI-SearchBot (ChatGPT search) and ChatGPT-User (user-triggered fetches); Perplexity’s PerplexityBot (search index) and Perplexity-User (user-triggered); Anthropic’s ClaudeBot (training), Claude-SearchBot (search quality) and Claude-User (user-triggered); Google’s Googlebot (Search, which feeds AI Overviews) and Google-Extended (a control for Gemini training and grounding); and Microsoft’s Bingbot. Each publisher documents these on its own site; check the current list before editing robots.txt.

Should we block GPTBot?

It depends on what you sell. OpenAI says disallowing GPTBot indicates that a site’s content should not be used to train its generative AI foundation models. Blocking it does not, by OpenAI’s description, remove you from ChatGPT search, which uses OAI-SearchBot. Publishers with valuable proprietary content often block GPTBot; a service business that wants to be recommended usually allows it, because model knowledge is one more route to being named. Decide deliberately and write the decision down.

What is OAI-SearchBot, and why must it be allowed?

OAI-SearchBot is the OpenAI crawler used, in OpenAI’s words, to surface websites in search results in ChatGPT’s search features. OpenAI states that sites which disallow OAI-SearchBot will not be shown in ChatGPT search answers. So if being cited in ChatGPT is a goal, this is the one user agent you must not block, regardless of what you decide about GPTBot. Check your robots.txt and any CDN bot rules for it explicitly.

What is the difference between PerplexityBot and Perplexity-User?

Perplexity describes PerplexityBot as designed to surface and link websites in search results on Perplexity, and states that it is not used to crawl content for AI foundation models. Perplexity-User, by contrast, fetches a page when a user asks a question, and Perplexity says that because a user requested the fetch, it generally ignores robots.txt rules. Allow PerplexityBot if you want to appear in Perplexity’s results; understand that user-triggered fetches may happen regardless.

What do ClaudeBot, Claude-SearchBot and Claude-User each do?

Anthropic documents three agents. ClaudeBot collects web content that could contribute to training its models. Claude-SearchBot navigates the web to improve search result quality for users. Claude-User accesses websites when individuals ask Claude questions. All three respect robots.txt, and Anthropic says site owners can also set a Crawl-delay for rate limiting rather than blocking outright. To be citable in Claude’s web search, allow Claude-SearchBot and Claude-User at minimum.

Does blocking Google-Extended remove a site from AI Overviews?

No. Google’s documentation says Google-Extended is a control publishers use to manage whether crawled content may be used for training future Gemini models and for grounding in Gemini apps and Vertex AI, and states that it does not impact a site’s inclusion in Google Search nor is it used as a ranking signal. AI Overviews and AI Mode are Search features fed by Googlebot. Blocking Googlebot would remove you; blocking Google-Extended would not.

How does Microsoft Copilot find web sources, and does Bingbot matter?

Copilot’s web answers draw on Microsoft’s search infrastructure, so in our working model Bingbot access and Bing indexing are the relevant controls; a site absent from Bing is unlikely to be cited by Copilot. Verify your site in Bing Webmaster Tools, submit a sitemap, and confirm Bingbot is not blocked by robots.txt or a CDN rule. Because Bing’s index covers many Indian sites less thoroughly than Google’s, this is often the quickest visibility win.

Can a user-triggered fetcher ignore robots.txt?

Yes, and the publishers say so. OpenAI says that because ChatGPT-User actions are initiated by a user, robots.txt rules may not apply. Perplexity says Perplexity-User generally ignores robots.txt for the same reason. These fetchers act like a browser on behalf of a person. Practically, robots.txt controls indexing and training crawlers, not live reads. If a page must never be read by an AI tool, it needs authentication, not a robots rule.

What should a robots.txt for AI search look like?

Keep it explicit. Allow Googlebot and Bingbot everywhere public. Add named entries for OAI-SearchBot, PerplexityBot, Claude-SearchBot and Claude-User with Allow: /. Decide separately on GPTBot, ClaudeBot and Google-Extended, and write the decision down. Disallow admin, staging, internal-search and cart paths for all agents. Point to your sitemap. Then test: fetch robots.txt, and check that the CDN or firewall does not override it with a block-all-bots rule that catches the agents you just allowed.

How do we know if AI crawlers are actually visiting the site?

Read the server or CDN access logs and filter by user agent: GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, ClaudeBot, Claude-SearchBot, Claude-User. Each publisher lists its user-agent string and IP ranges. Count fetches per week and which paths are read. If a crawler you allowed never appears, check for a CDN bot-management rule or a firewall block. If it fetches only the home page, your internal linking or sitemap is not exposing the deeper pages.

How do structured data and entity consistency affect AI search?

Does schema markup help with AI search optimisation?

Indirectly. Google states there is no special schema.org structured data you need to add to appear in AI Overviews or AI Mode. But schema is how Google and others confirm what an entity is, and consistent Organization, Article and FAQPage markup reduces ambiguity about your name, location and founding date. Treat schema as entity hygiene rather than an AI ranking lever: keep it accurate, matching the visible text, and identical in meaning to your LinkedIn and Google Business Profile.

Which schema types matter most for a service business?

Four cover most cases. Organization (with @id, address, foundingDate and sameAs links to LinkedIn and social profiles) on the entity page. Service for each service line, with provider pointing to the Organization. Article with an Organization author for journal posts. FAQPage for question hubs. BreadcrumbList helps engines see site structure. Everything in the markup must be visible on the page; hidden schema-only facts are a policy risk and add no trust.

What is entity consistency, and why do AI engines care?

Entity consistency means every public source states the same core facts about you: legal name, trading name, city, founding year, what you do, who you serve, phone and email. AI engines reconcile sources before they name a business; contradictions make them hedge, generalise or skip you. Audit your site, LinkedIn, Google Business Profile, directories and press, and fix every mismatch. Yamm Labs repeats one line everywhere: a design-led brand agency for fintech and BFSI companies, founded in Gurugram in 2017.

Should the About page or the home page be the entity home?

Either can, but pick one and make it unambiguous. The entity home is the page you want engines to treat as the canonical description of your organisation: it carries the full Organization schema with @id, the disambiguation line if your name collides with another brand, the founding year, address and client list. Most agencies use About. Link to it from every page footer and from LinkedIn, and reference its @id from the schema on every other page.

Does FAQPage schema still matter now that Google shows fewer FAQ rich results?

For rich results, Google has reduced how often FAQ markup produces a visible SERP feature, so that benefit is small. For AI search, FAQPage schema still helps engines parse question and answer pairs cleanly, and it keeps the machine-readable version identical to the visible text. The larger benefit is the format itself: a question heading followed by a self-contained 50 to 90 word answer is exactly what AI engines quote. Publish the FAQ for the format, add the schema for hygiene.

How does Google’s Knowledge Graph relate to AI answers?

The Knowledge Graph is Google’s database of entities and their relationships; a Knowledge Panel is its visible form. AI Overviews, AI Mode and Gemini draw on it to identify who a company is, so an accurate Knowledge Panel makes it more likely that generated answers describe you correctly. You influence it through consistent schema, a verified Google Business Profile, Wikidata where eligible, and third-party sources that repeat your facts. Claim the panel if Google offers it.

Do author pages and bylines matter for AI search if the author is the organisation?

An organisational byline works if the organisation is a clear entity. AI engines look for who is responsible for a page; a byline linked to an entity page with Organization schema, contact details and a history satisfies that. Named individual authors add credibility for opinion and expert content, but they are not required. Yamm Labs publishes under the organisation’s name with an Organization author in Article schema, and keeps the entity page as the accountability anchor.

Must the company name be identical on LinkedIn, Google Business Profile and the website?

Yes, as far as the platforms allow. Use one trading name everywhere, in the same spelling and capitalisation, and put the legal entity name only where it is required. If your name collides with another brand, add a disambiguation line on your entity page and in the LinkedIn About text. Yamm Labs, for example, states that it is not related to Yet Another Mail Merge (YAMM) or to yaM Labs, so engines do not merge the three.

How do we fix a wrong fact about our company in a Knowledge Panel or AI answer?

Find the source. Ask the engine to cite it, or search the wrong fact in quotes. Correct it at the origin: your own page, a directory, a press article (ask for a correction) or Wikidata. Then make sure the correct fact appears on your entity page with schema and on two or three corroborating sources. For a Knowledge Panel, use Google’s suggest-an-edit feature once you have claimed it. Re-query after two to four weeks.

Does the postal address and phone number really need to be on the website?

For a business that wants to be recommended, yes. An address and a working phone number are among the simplest legitimacy signals an engine can verify against Google Business Profile and directories. Put them on the contact page, in the footer and in Organization schema, in the same format everywhere. If you work remotely, use a registered office address rather than omitting it. A business with no findable location is treated as less certain.

Which content formats get cited by AI engines?

What is an example of answer engine optimisation done well?

A clear example is a page whose H1 is the question buyers type, followed by a 60 to 100 word direct answer, a table of contents, and sections whose headings are the follow-up questions. Each section restates the noun and answers in the first sentence. The page carries a visible update date, an organisational byline and a Sources list. Our GEO and AEO FAQ and this page are built on that pattern, so an engine can lift any section intact.

How should you write a paragraph so an AI engine can quote it?

Lead with the answer, name the subject in the first sentence, and keep the paragraph to 50 to 90 words. Use concrete nouns, numbers and dates rather than adjectives. Avoid pronouns that depend on the previous paragraph, because the passage may be lifted alone. End with a rule of thumb or an example. If the paragraph would be understood by someone who read nothing else on the page, it is ready.

Do statistics, quotes and citations increase the chance of being cited?

In our experience, and in the GEO research paper published in 2023, yes: that paper reported visibility gains of up to 40 percent in generative engine responses from content changes such as adding statistics, quotations and citations. Passages that carry a specific number, a quoted authority or a cited source give the engine something verifiable to repeat. The condition is that the numbers are real and attributed; invented figures are a liability, especially for regulated brands.

Do question-phrased headings help?

Yes. A heading that matches how people ask (“How much does fintech branding cost in India?”) gives the retrieval step an exact lexical match to the sub-query, and tells the model that the section below is the answer. Use one question per heading, keep it under twelve words, and do not stack several questions into one section. This is the simplest structural change most sites can make, and it is why question hubs like this one exist.

Should we publish definition and glossary pages?

Yes, if your buyers use jargon. “What is X” prompts are among the most common in AI search, and engines like a short, sourced definition they can attribute. Publish one page per important term, or a glossary with anchor links, each entry 50 to 90 words with the term in the first sentence. Link each definition to the service page it relates to. For BFSI, terms like NFO, PSW or GEO are good candidates.

How long should an answer be for AI search?

Long enough to be complete, short enough to be lifted. In practice, 50 to 90 words per section for FAQ answers, and 60 to 120 words for the short answer at the top of a page. Longer sections are fine when they are broken by sub-headings so that each part can stand alone. The page as a whole can be long; the unit that gets cited is the passage, so make every passage complete.

Do images, video and audio help with AI search?

Sometimes. Engines increasingly index video and image content, and video transcripts can become citable text. For most service businesses the practical steps are a transcript or summary under every video, descriptive alt text, and captions that state the fact shown. Do not rely on an image to carry a key fact; write it in the text too, because text is what engines retrieve first. Our fintech explainer video work always ships with a written summary for that reason.

Should we update old posts or write new ones?

Update first. An old post that already ranks and already has links is a better vehicle for AI citation than a new page with no history. Rewrite the opening as a direct answer, convert headings to questions, add a table, correct facts, and update the visible date. Then write new pages only for questions the site does not yet answer. A quarterly pass over the twenty most important pages beats fifty new posts.

Does the URL slug or page title affect AI citations?

Modestly. The title is often what an engine shows as the citation label, so a title that states the answer reads better than a clever one. The slug should be short and descriptive. Neither is a strong retrieval signal on its own, but both help the engine and the reader recognise what the page is for. Match the title, the H1 and the first sentence in meaning, so the citation label and the quoted passage agree.

How do we optimise an existing page for AI search without rewriting it?

Do six things. Add a 60 to 100 word short answer under the H1. Convert the main headings to questions. Add a table of contents with anchors. Make the first sentence of each section a direct answer. Add a visible update date and a Sources list. Add FAQPage or Article schema that matches the text. Leave the body prose intact. This takes an afternoon per page and captures most of the gain.

Why do reviews, listicles, Reddit and other third-party sources matter so much?

Can a business control its Google reviews?

No, and attempting to is against Google’s policies. A business can respond to reviews, flag ones that breach policy, and ask real customers for honest reviews. It cannot delete negative reviews, buy reviews, or gate requests so that only happy customers are asked. For AI search, a steady flow of genuine reviews that mention specific services is more useful than a perfect score, because engines read the text for what you actually do.

How should we ask customers for reviews that help AI visibility?

Ask promptly after a completed project, by a personal message with a direct link, and suggest the customer mention what was done and for whom. Never script the review or offer an incentive. Specific text is what AI engines can use; a five-star rating with no words tells them nothing. Spread requests across Google, LinkedIn recommendations and one relevant directory rather than piling everything on one platform, so that several sources corroborate each other.

Are backlinks still needed for AI search?

Yes, but for a different reason. Backlinks still drive the rankings that supply candidate pages to AI engines, and a page nobody links to is rarely retrieved. But for AI answers, the mention matters as much as the link: an unlinked reference on a trusted page still corroborates your entity. Prioritise mentions on pages engines already cite for your category, whether or not they link, and keep the mention consistent with your entity line.

Do YouTube videos and podcasts get cited by AI engines?

Increasingly. YouTube is indexed by Google and transcripts are readable, so a video that answers a question clearly can appear in AI Overviews and AI Mode. Podcast episodes are cited when the host publishes show notes or a transcript on a crawlable page. The practical rule: whatever is said on audio or video should also exist as text on a page you or the host control, with your entity name spelled consistently.

Do awards and certifications help a business get recommended?

Only if they are verifiable. An award listed on the awarding body’s own site is third-party corroboration; an award badge only on your site is a claim. Certifications work the same way: a certificate number that can be checked, or a directory badge that links to the certifier. Never invent or exaggerate; engines cross-check, and regulated brands face compliance risk for unverifiable claims. If it cannot be checked, leave it out.

Do guest posts and contributed articles still count?

Yes, when they are on sites engines already cite and when they are genuinely useful. A contributed article on a trade publication that describes your expertise, names your firm once and links to your entity page is a strong corroborating source. Mass guest posting on low-quality blogs is not; engines weight the host. Aim for two or three placements a year on publications your buyers read, and write them to be quotable.

How do we get mentioned on Reddit without it looking like spam?

Participate as a person, not as a brand. Answer questions in the relevant subreddits with useful detail, disclose your affiliation when it is relevant, and mention your company only when someone asks for a recommendation and it genuinely fits. Never create fake accounts or ask staff to upvote. Reddit threads are heavily cited by AI engines because they read as unfiltered opinion, which is exactly why manufactured ones are detected and damage trust.

Do case studies published on a client’s or partner’s site help?

Yes. A case study hosted by the client or a partner is a third-party source that names you, describes the work and often links back. It is more valuable than the same case study on your own site because the engine treats it as corroboration rather than a claim. Ask clients, within your confidentiality terms, for a short published note or a LinkedIn post naming the work. Keep outcome claims to what the client will state publicly.

Do B2B directories like G2, Clutch or Sortlist matter for an agency?

They matter as sources, because engines cite them when answering “best agency for X” prompts. A complete, accurate profile with services, location, founding year and real reviews is worth maintaining on two or three directories your category uses. Paid tiers buy placement on the directory, not in the AI answer, so weigh them like any advertising. See the note on directories in our GEO and AEO FAQ for how we treat them.

Should we respond to negative mentions that AI engines might quote?

Yes. A negative review or thread on a cited source will be paraphrased into answers. Respond factually and calmly on the same platform, correct errors, and where a complaint was resolved, say so; engines read the reply too. Do not argue or threaten. If a mention is defamatory, use the platform’s process. Over time, add positive, specific third-party sources so the negative one becomes one voice among many rather than the only one.

How do you measure AI visibility, and which tools help?

How do you measure generative engine optimisation?

Measure it as share of answer: for a fixed set of prompts, across a fixed set of engines, how often your brand is Present, Partial or Absent, and which sources are cited. Record it monthly. Add accuracy (are the facts about you correct), sentiment (how you are described) and source count (how many distinct third-party pages engines cite for you). Referral traffic from AI engines is a lagging, partial indicator; the answer itself is the primary metric.

What are AI search optimisation tools, and what do they do?

AI search optimisation tools automate the prompt-and-record loop: they run a set of prompts against ChatGPT, Perplexity, Gemini and Google’s AI features, log whether your brand and competitors appear, capture the cited sources, and chart change over time. Well-known examples include Profound, Peec AI, Otterly and Semrush’s AI toolkit. Some add sentiment analysis and content recommendations. None can place you in an answer; they measure and prioritise, and you still do the work.

How do we compare AI search optimisation tools?

Compare on six points: which engines and countries they query; whether they run prompts fresh or use cached data; how many prompts and competitors are included per plan; whether cited sources are captured as URLs; whether you can export raw data; and price relative to your prompt volume. Run the same 20 prompts manually first, then trial two tools against your manual baseline. If a tool disagrees sharply with what you see by hand, ask why before buying.

Can we track visits from ChatGPT and Perplexity in Google Analytics?

Partly. Referral traffic from ChatGPT, Perplexity, Copilot and Gemini usually arrives with an identifiable referrer, so you can build a channel group in GA4 that matches those hostnames. AI Overviews and AI Mode are reported as Google organic and cannot be separated in GA4 today. Expect small absolute numbers; the value of AI visibility is mostly brand recall and shortlist inclusion, which show up later as branded search and direct enquiries.

What is a prompt set, and how big should it be?

A prompt set is the fixed list of questions you run every month to measure visibility. Twenty to forty prompts is enough for one service line: category, problem, comparison and location prompts in roughly equal share, written the way buyers type them. Keep it stable so month-on-month change is meaningful; add new prompts in a separate group. Yamm Labs starts BFSI clients with 30 prompts across four engines and reports Present, Partial or Absent per prompt.

Do AI answers change by user, location or time of day?

Yes. Engines personalise on account history, location and language, and generated answers vary run to run even for the same prompt. So one screenshot proves little. Measure with a logged-out or fresh session, a fixed location setting and three runs per prompt, then record the majority outcome. Tools handle this by repeating prompts. When a client says an engine does not mention them, ask how it was tested before changing anything.

What is the difference between a citation and a mention in an AI answer?

A citation is a link to your page shown as a source; a mention is your brand named in the answer text, with or without a link. Both matter. A mention without a citation usually means the engine took your name from a third-party source; a citation without a mention means your page supplied a fact but the answer did not recommend you. Track both, and note which of your pages, or which third-party pages, are cited.

Can Perplexity be used as an SEO or research tool?

Yes, usefully. Because Perplexity shows its sources, you can run buyer prompts and immediately see which pages the engine trusts for your category, which is a ready-made list of corroborating sources to pursue. You can also ask it to summarise how your brand is described and where that description came from. Treat the output as leads to verify, not as facts; it is a research assistant, not a ranking tool.

Is there a free way to monitor AI visibility?

Yes, at small scale. Keep a spreadsheet of 20 prompts, run them monthly by hand in ChatGPT, Perplexity, Gemini and Google AI Mode in a fresh session, and record Present, Partial or Absent plus the cited URLs. It takes two to three hours a month. Set up GA4 referral grouping for AI hostnames. Search your brand name in quotes on Reddit and news monthly. Move to a paid tool when the prompt set outgrows the spreadsheet.

What should an AI visibility report contain?

Five parts. The prompt set and method. A Present, Partial or Absent grid by prompt and engine, with month-on-month change. The cited sources, split into your pages and third-party pages. A fact-accuracy check with any errors and their origin. A ranked fix list: what to correct, what to publish, which third-party sources to pursue. Keep it to two pages; the grid and the fix list are what a CMO actually reads.

How does AI search optimisation work for BFSI and regulated brands?

Why is AI search optimisation different for banks, insurers and AMCs?

Because every sentence an engine might paraphrase is subject to SEBI, IRDAI or RBI advertising rules, and the engine will not carry your disclaimer with it. A fund page that implies returns, an insurance page that overstates a benefit, or a lending page that hides a charge can be quoted without context. BFSI teams must therefore write pages that are compliant when lifted sentence by sentence. Our AI visibility service for BFSI is built around that constraint.

Can a regulated brand do GEO without breaching advertising rules?

Yes, if the content is factual, sourced and free of projections, testimonials and superlatives, which is what the advertisement codes require anyway. Describe products by their stated objective, category and features as filed. Put the standard warnings in visible text near the facts, not only in a footer. Have compliance review the answer-first sections specifically. Confirm the current circular with your compliance team, because SEBI, IRDAI and RBI have all updated rules recently.

Which pages of a bank or AMC are most likely to be cited?

Explainer and FAQ pages rather than product landing pages. Engines cite pages that answer “what is”, “how does”, “which is better” and “is it safe” questions in plain language: how an NFO works, what a risk-o-meter means, how a claim is settled, what a loan’s charges are. Product pages full of adjectives and buttons are rarely quoted. Build a question hub per product category, keep it factual, and link it to the product page.

Should a BFSI brand allow AI crawlers at all?

For public marketing and education pages, yes; being absent from AI answers is a bigger risk than being quoted, because competitors and third parties will fill the gap. Block crawlers from customer portals, transactional paths, internal search and anything behind login, and use authentication rather than robots.txt for genuinely private content. Decide on training crawlers (GPTBot, ClaudeBot, Google-Extended) separately with legal, and document the decision.

What is the mis-selling risk when an AI engine paraphrases our page?

The risk is that the engine drops a condition. “Returns up to 12 percent” becomes “returns 12 percent”; “subject to underwriting” disappears. You cannot prevent paraphrase, so remove the material that is dangerous when shortened: no projections, no headline numbers without the qualifying sentence in the same paragraph, no benefit stated without its main exclusion. Re-query your product prompts monthly and correct the source page if an answer is misleading.

How should disclaimers and standard warnings be handled for AI search?

Put them in the passage, not only at the page foot. If a paragraph states a fund’s category and objective, the next sentence carries the standard warning in plain text. Keep the wording exactly as the regulator prescribes; do not shorten it for style. In llms.txt or schema, do not attempt to encode disclaimers; engines will not honour them there. Our BFSI marketing compliance FAQ covers the SEBI, IRDAI and RBI wording rules in detail.

Should an AMC publish NFO pages for AI engines?

Yes, as factual explainers within the NFO period rules. An NFO page that states the scheme name, category, objective, benchmark, risk-o-meter and the standard warning, with the dates and where to read the SID, is exactly what an engine cites for “what is the X NFO” prompts. Avoid performance of similar funds and anything that implies returns. Prepare the page before the window opens so it is indexed on day one. See our NFO launch communication page.

Can an insurer publish claim-settlement or solvency figures for AI engines to quote?

Only figures that are already public disclosures, stated exactly as filed, with the period and source. Public disclosure data, when reproduced accurately and dated, gives engines something verifiable, and being the source of your own numbers is better than leaving it to comparison sites. Do not round favourably, do not compare against competitors, and confirm with compliance how the figure may be used in a promotional context before publishing.

What does an AI visibility agency actually do for a BFSI brand?

It runs the measurement loop and fixes the sources. In our case: baseline 30 prompts across four engines, audit entity consistency and crawler access, restructure the pages engines should cite, write compliant answer-first explainers, pursue third-party corroboration, correct wrong facts at their origin, and re-query monthly with a fix list. Everything goes through the client’s compliance team. What it cannot do is guarantee placement. See our AI visibility and GEO agency for BFSI page.

Where should a BFSI marketing team start with AI search optimisation?

Start by asking the four engines your ten most important buyer prompts and screenshotting the answers. Note where you are absent, where facts are wrong, and which sources are cited. Then fix entity consistency and robots.txt, rewrite the three pages most likely to be cited as answer-first explainers, and pursue two third-party sources. Re-query in a month. That first cycle costs a few days and tells you whether a larger programme is worth it. Talk to Yamm Labs if you want help.

Want to know whether ChatGPT, Perplexity, Gemini and Google AI Mode name your brand, and what they say?

Yamm Labs runs a 30-prompt baseline across four engines, fixes entity and crawler issues, restructures the pages engines should cite, and re-queries monthly with a compliance-reviewed fix list. See the AI visibility and GEO agency for BFSI page, or Talk to Yamm Labs →

Last updated: 18 September 2026

Sources

  1. Overview of OpenAI Crawlers (OAI-SearchBot, ChatGPT-User, GPTBot), OpenAI, accessed 18 September 2026
  2. Perplexity Crawlers (PerplexityBot, Perplexity-User), Perplexity, accessed 18 September 2026
  3. Does Anthropic crawl data from the web, and how can site owners block the crawler? (ClaudeBot, Claude-User, Claude-SearchBot), Anthropic, accessed 18 September 2026
  4. Google’s common crawlers (Googlebot, Google-Extended), Google Search Central, accessed 18 September 2026
  5. AI features and your website (AI Overviews and AI Mode), Google Search Central, accessed 18 September 2026
  6. The /llms.txt file, Jeremy Howard, first published 3 September 2024, v2 10 August 2026
  7. AGENTS.md: a simple, open format for guiding coding agents, Agentic AI Foundation (Linux Foundation), accessed 18 September 2026
  8. GEO: Generative Engine Optimization, arXiv, submitted 16 November 2023, revised 28 June 2024
  9. GEO and AEO FAQ: 60 questions for BFSI brands, Yamm Labs, updated 18 September 2026
  10. AI Visibility and GEO Agency for BFSI, Yamm Labs, updated 18 September 2026