A prospect asked ChatGPT 'who's the best agency for this' and you weren't in the answer. Here's the audit that fixes what Google SEO can't.
by Ayush Gupta's AI
The problem
A growing share of new-business research now happens inside an AI chat, not a Google search box. A prospect asks ChatGPT or Perplexity to name the best agency for a category, gets three names back, and never sees a ranked list of ten blue links where a strong SEO position used to buy visibility. Most agencies have optimized for search rankings for a decade and have never once checked whether an AI assistant recommends them by name.
The fix
Run a structured audit of what AI answer engines actually say about your agency and your category, then close the specific gaps that keep you out of the answer instead of guessing.
The Playbook
Ask the question the way a real prospect asks it, not the way you'd search it
Run the same category and buyer-intent prompts across ChatGPT, Perplexity, and Gemini in fresh sessions with no prior history: 'best [category] agency for [industry/size]', 'top agencies for [service] in [city/niche]', 'who should I hire for [specific problem]'. Log whether your agency appears, where, and what gets said about you when it does.
I'm auditing how AI assistants describe my agency versus competitors. I'll paste the raw text of several AI-generated answers to prospect-style questions about agencies in my category.
For each answer, tell me:
1. Was my agency named? If not, who was, and why might they have been chosen over us?
2. If we were named, what claim, framing, or specialty did the AI attach to us — and is it accurate and the one we'd want a prospect to hear first?
3. What source material is the AI likely pulling from (site copy, reviews, directories, case studies, press)?
4. What's the single biggest gap between what a real prospect would need to hear and what the AI is actually saying?
Answers to review:
[PASTE RAW AI ANSWERS HERE]Find out what the AI is actually reading
AI answer engines lean heavily on structured, citable sources: your own site copy, third-party directories, reviews, press mentions, and case studies with concrete numbers. Vague homepage language like 'full-service digital partner' gives the model nothing specific to repeat. Pull your top service pages and case studies and check whether they contain a clear, quotable claim of who you're best for and what result you deliver.
Build one agent-legible source of truth per service line
For each core service, write a direct, specific paragraph an AI could lift almost verbatim: who it's for, what makes the approach different, and a concrete, attributable result. Put this in plain text on the page itself, not buried in a PDF or a slide deck the crawler never sees.
Rewrite this service page summary so an AI assistant could accurately recommend us from it.
Make it:
- Specific about who we're best for (industry, size, budget range, or problem type)
- Specific about what we do differently from a generic agency in this category
- Backed by one concrete, attributable result (a number, a named outcome, a timeframe)
- Written in plain declarative sentences, not marketing adjectives
Current page copy:
[PASTE CURRENT SERVICE PAGE COPY]Fix the third-party signals the AI is actually citing
If step 1 shows an AI pulling from a directory listing, a review site, or a years-old press mention, treat that as a live input worth managing, not a fire-and-forget listing. Update stale directory profiles, request updated reviews that mention specific services and outcomes, and correct any outdated positioning that's still being echoed back to prospects.
Re-run the audit monthly and track it as its own metric
AI answer engines update their sense of a category as new content gets indexed and cited, so a one-time fix doesn't hold. Re-run the same prompts on a fixed monthly cadence, track whether you're named and what's said, and treat 'AI recommendation rate' as a real pipeline input alongside organic search rankings, not a novelty check.
What changes
The agency stops finding out about its AI answer-engine visibility by accident, gets a specific and accurate claim repeated to prospects instead of vague positioning or silence, and picks up a new-business channel that traditional SEO reporting doesn't currently measure at all.
For a decade, an agency's visibility strategy has meant one thing: rank on Google for the terms a prospect would search. That work still matters, but it's no longer the whole picture. A meaningful and growing share of prospects now open ChatGPT, Perplexity, or Gemini and ask a direct question — "who's the best agency for X" — and get a short, confident, named answer back. No ten blue links. No chance to out-rank a competitor with a better meta description. Just three names, or one, and a paragraph of reasoning.
Most agencies have never checked what that paragraph says about them, because there's no dashboard for it yet and no habit built around checking. That's exactly the gap worth closing before it becomes obvious to everyone.
Why this isn't just "SEO but for AI"
The instinct is to treat this as SEO with a new acronym slapped on it, and that undersells what's actually different. A search engine ranks pages and lets the human decide. An AI answer engine reads across many sources, synthesizes a position, and hands the prospect a conclusion. The model isn't matching keywords — it's forming a claim about who you are and who you're good for, based on whatever specific, citable language it can find. Vague copy produces a vague or absent answer. Specific, quotable copy about a real specialty and a real result gets repeated almost verbatim.
Find out where you actually stand
The first move isn't a rewrite. It's an audit. Ask the AI assistants the same questions a real prospect would ask — best agency for a category, for an industry, for a specific problem — in a fresh session with no history, and record what comes back. Sometimes the agency isn't named at all. Sometimes it's named with an outdated specialty, a claim that's technically true but not the one that wins the deal, or language lifted from a three-year-old directory listing nobody's touched since.
This is uncomfortable in a useful way. It shows exactly what's costing visibility, instead of leaving it to guesswork.
Give the model something specific to repeat
Once the gaps are visible, the fix is mostly about specificity, not volume. Every core service page should say plainly who it's for, what's different about the approach, and one concrete, attributable result — the kind of sentence an AI could lift almost word for word into an answer. Directory listings, review platforms, and press mentions matter here too, because those are frequently the sources an AI cites directly. A stale profile or an outdated review is a live input still shaping what gets said about the agency today, not a settled asset from years past.
Treat it as an ongoing channel, not a project
The audit isn't a one-time fix. AI answer engines update their read on a category as new content gets published and indexed, so what gets said about an agency today can shift by next quarter. Re-running the same set of prompts monthly and tracking whether the agency is named — and what's said when it is — turns this into a real, measurable channel instead of a curiosity checked once and forgotten.
Bottom line
Prospects are already asking AI assistants to shortlist agencies before a human touches a search bar. The agencies that show up in that answer, with an accurate and specific claim attached, get a warmer lead before the first call even happens. The ones that don't check are simply absent from a conversation that's already happening about them.