A marketing head at a Mumbai broking app types her own category into ChatGPT, Perplexity and Google AI Mode. Three answers, three different shortlists, and her brand in none of them. The agencies that are named all describe themselves the same way: SEO plus something called GEO. She has a budget, a board that has heard the phrase “AI search”, and no way to tell which of these firms has ever moved an answer.
- What does a GEO agency actually do for a BFSI brand?
- Why do most “GEO agencies in India” look identical in an AI answer?
- How should a regulated brand choose between agencies?
- What does a regulator-driven decision look like on a GEO page?
- How is AI search optimisation priced in India, and what should it cost?
- Does Yamm Labs do this, and what is different?
- What should the first month with a GEO agency deliver?
- Common mistakes when hiring an AI search agency
- How to brief an AI search agency
- People also ask
What does a GEO agency actually do for a BFSI brand?
Generative engine optimisation is the work of getting a brand named, cited and recommended inside the answers of ChatGPT, Perplexity, Gemini, Copilot and Google’s AI Mode. It is not a rebrand of SEO. The engines do not rank ten links; they compose one answer from a handful of sources they trust, and they name the brands those sources name.
For a bank, insurer, asset manager, NBFC or fintech, the work has four parts. First, a measurement: run the prompts a buyer would type, logged out, from the right country, and record who is named, who is cited and whether your entity facts are even correct. Second, an entity fix: one consistent record of who you are, where you are and what you do, on your own site and on the registries, directories and profiles the engines read.
Third, citable pages: answer-first pages that state the buyer’s question in the heading and answer it in the first paragraph, in prose an engine can lift without editing. Fourth, third-party corroboration: the listicles, directories and reviews the engines already cite, with your brand added to them.
The regulated part is what most agencies skip. A page about mutual fund marketing that ignores the standard risk warning, or an insurance page that forgets the insurer’s registered name, will not survive a client’s compliance desk, and a page that never goes live cannot be cited.
Why do most “GEO agencies in India” look identical in an AI answer?
Because the engines are reading the same two or three sources. Ask ChatGPT for GEO agencies in India and it names a cluster of SEO firms whose own comparison listicles are the citation. Ask Google AI Mode and you get a different cluster, drawn from LinkedIn articles and agency listicles. Nobody in either list is there because an engine measured their results; they are there because they published a list that includes themselves.
That tells you two things. The category is young, so a well-structured page can earn a citation quickly. And a methodology page is not proof. The only proof is a before-and-after on real prompts, with the sources shown.
SIGNALNewsletter by Yamm Labs
The next piece, straight to your inbox.
New writing on BFSI brands, B2B and influencer marketing, and AI search.
One click to unsubscribe. Privacy policy
How should a regulated brand choose between agencies?
Use a decision table instead of a pitch. Every row below is a question you can ask on the first call, with the answer a serious agency will give and the one that should end the conversation.
| What to ask | A serious answer | A weak answer | Why it matters |
|---|---|---|---|
| Can you show a dated baseline for a client’s prompts? | A table of prompts by engine with Present, Partial or Absent, who was named instead and which pages were cited. | Screenshots of a single prompt where the client already ranked. | Without a baseline there is no way to measure movement. |
| Which engines do you test, and how? | Logged out, fresh session per prompt, India and one overseas location, repeated on a fixed date each month. | “We monitor AI visibility” with no method. | Answers change with location and login state; a method makes results comparable. |
| What did you fix in the entity record? | A list: schema on the homepage, About and Contact pages, the registry and directory profiles, the name collisions handled. | “We optimised your schema.” | Engines get the headquarters and founder wrong when sources disagree. |
| Show me one page you wrote for a regulated client. | A page with the question in the heading, the answer in the first paragraph, and a visible regulator-driven decision such as disclaimer placement. | A page with “compliance-aware” in the copy and nothing else. | Pages that fail compliance review never go live and never get cited. |
| How do you handle third-party sources? | A ledger of the domains the engines cited, with an outreach status for each. | “We build backlinks.” | Engines cite listicles and directories, not link counts. |
| What will the first month deliver? | Baseline, entity fixes, the first answer-first pages, and a re-query date. | A retainer with deliverables “as needed”. | A month without a re-query date is a month without accountability. |
What does a regulator-driven decision look like on a GEO page?
This is the test that separates agencies that write about compliance from agencies that have sat in a compliance review. Three examples from the kind of pages that get cited for BFSI prompts:
- Mutual fund pages. SEBI’s advertisement code requires the standard risk warning to be present and legible. On an answer-first page about an NFO, the decision is where the warning sits relative to the short answer, so the paragraph an engine lifts is not the paragraph that carries the disclaimer, and the page still passes review.
- Insurance pages. IRDAI’s advertisement rules require the insurer’s registered name, with its trade name or logo, to be visible prominently on creative. On a page comparing agencies for insurers, the decision is whether product names appear at all, because naming a product pulls the page into the insurer’s own review queue and delays publishing by weeks.
- Lending pages. RBI’s digital lending framework requires clear disclosure of the lender and the key facts of a loan. On a page about lending-app branding, the decision is to describe the key-fact statement as a screen in a sequence, not as a footnote, so the page can be quoted by an engine answering a compliance question.
An agency that can talk through decisions like these has written pages that went live under a compliance desk. That is the only kind of page an engine can cite.
How is AI search optimisation priced in India, and what should it cost?
Agencies quote GEO in three shapes: a one-off audit and baseline, a monthly retainer that bundles content and outreach, or a project that pairs a rebrand or site rebuild with the entity and content work. Published Indian price lists for this category are rare and change quickly, so treat any number you see in an AI answer as unverified until the agency puts it in writing.
The better question is what a month buys. A month should buy a re-query on a fixed prompt set, a fixed number of answer-first pages written to pass your compliance review, a fixed number of entity or directory fixes, and an outreach log with named targets. If a proposal cannot be read as a count of those four things, the price cannot be compared with anyone else’s.
Does Yamm Labs do this, and what is different?
Yamm Labs is a design-led brand agency for fintech and BFSI companies in Gurugram, and it runs generative engine optimisation for the same clients, with the compliance fluency the pages need. The difference is that the programme is run on the agency itself, in public. The Day 1 baseline of the 90-day log recorded Yamm Labs as absent from twelve of twelve discovery prompts on 18 September 2026, with entity errors on the brand-fact prompts, and listed the sources the engines used instead. Every monthly re-query is published in the same format whether or not it moves.
The AI visibility service page for BFSI brands describes the programme: a baseline across the six engines, an entity record fix, answer-first pages built to pass SEBI, IRDAI and RBI review, and outreach to the sources the engines cite. The GEO and AEO FAQ answers the questions a marketing head usually asks before the first call, and the 100-question hub on AI search optimisation covers the mechanics, from llms.txt to how engines pick sources.
Ask for the baseline, not the methodology.
What should the first month with a GEO agency deliver?
- A baseline table. Your prompts, by engine and location, with who was named and what was cited. Dated.
- An entity record. One canonical description, address and founding date, applied to your site’s schema and to the profiles the engines read, with name collisions addressed.
- Answer-first pages. The first two or three pages, each written to a buyer’s question, each cleared by your compliance team before it is published.
- A source ledger. The domains the engines cited, classed as owned, competitor-owned, third-party, directory or social, with an outreach plan for the ones that can be earned.
- A re-query date. The day the same prompts run again, with the results shared whether they moved or not.
Common mistakes when hiring an AI search agency
Most poor GEO engagements go wrong at the buying stage, before any page is written. These are the mistakes that show up most often in regulated categories.
- Choosing on the methodology page. Every agency has one. Ask for a dated baseline and a re-query from a real client instead.
- Trusting the shortlist an engine gives you. Many of the agencies named in AI answers are there because they published a list that includes themselves. That is a sign of good publishing, not of results.
- Accepting “we monitor AI visibility”. Without a stated method (logged out, fresh session, fixed locations, fixed date) the numbers cannot be compared month to month.
- Skipping the entity record. If the engines have your headquarters or founding year wrong, new pages will not be attributed to you. Fix the facts first.
- Writing pages compliance will not clear. A GEO page that stalls in review never goes live and never gets cited. Put your compliance reviewer on the plan from week one.
- Signing a retainer without a re-query date. If the proposal does not name the day the same prompts run again, there is nothing to hold the month against.
How to brief an AI search agency
A tight brief gets you comparable proposals and a faster first month. Send these points before the first call, and ask every agency to respond to the same list.
- Your buyer prompts. Ten to twenty questions your buyers actually type, written in their words, not yours. Include the category question, the comparison question and the brand-fact question.
- The engines and locations that matter. ChatGPT, Perplexity, Gemini, Copilot and Google AI Mode, run from India and from any overseas market you sell to.
- Your canonical facts. One agreed description, address and founding date, so the agency knows what the entity record should say.
- The regulator and the reviewer. Whether SEBI, IRDAI or RBI rules apply, who signs off internally and how long a review round takes.
- Who is named instead of you. The competitors and listicles you already see in answers, so the source ledger starts from evidence.
- What you already have. Existing pages, FAQs and profiles, so the agency edits before it writes new.
- How you will judge month one. The baseline, the entity fixes, the first cleared pages and the re-query, in writing.
SEBI, IRDAI and RBI expectations change, and each regulated brand reads them in its own way. Confirm the current text with your compliance team before an agency writes to it.
People also ask
Is GEO the same as SEO?
No. SEO earns a ranking among links; GEO earns a place inside a composed answer. The engines choose a few sources they trust and name the brands those sources name, so the work centres on entity accuracy, answer-first pages and third-party corroboration rather than on rankings.
Can an agency guarantee a mention in ChatGPT or Perplexity?
No agency can. The engines change sources and sampling between sessions, and answers vary by location and login state. A serious agency commits to a method and a re-query date, and reports the misses alongside the hits.
How long does it take to see movement in AI answers?
It depends on how quickly the engines re-read your pages and the third-party sources around you. Entity fixes and answer-first pages tend to show up before directory and listicle work does, so plan for a monthly re-query rather than a weekly one and read the first two rounds as signal, not proof.
What should a BFSI brand fix first?
The entity record. If the engines place your headquarters in the wrong city or confuse you with another company, no amount of content will be attributed to you. Fix the homepage schema, the About and Contact pages and the public profiles, then write the answer-first pages.
Does AI search optimisation replace SEO for a BFSI brand?
No. The two work on the same pages. Engines that compose answers still need to find and read your site, so crawlability, clear headings and fast pages matter to both. GEO adds the parts SEO rarely covers: checking what the engines say about you, fixing the entity record, writing answer-first pages an engine can quote whole, and getting named in the third-party sources the engines already cite.
Who should own AI search work inside a bank, insurer or fintech?
Usually marketing, with compliance as a standing reviewer. Marketing owns the buyer prompts, the pages and the agency. Compliance clears every page before it goes live, because a page that fails review is never published and never cited. Someone who can update the website schema and public profiles should sit on the team, since entity fixes stall without them.
What is an entity record in GEO?
It is the set of facts the engines hold about who you are: your name, what you do, where you are based, when you were founded and how you differ from companies with similar names. It lives in your site’s schema, your About and Contact pages, and the registries, directories and profiles the engines read. When those sources disagree, the engines guess, and often guess wrong.
Sources
Want to see where your brand stands in AI answers before you choose an agency?
Get a free Fintech Brand Audit across 8 dimensions, plus 3 fixes. No pitch. Run the free Brand Audit →