Ask ChatGPT, Claude and Perplexity the five buyer questions agencies actually get asked, then read the answer literally. What matters is not whether your name shows up anywhere, but whether you are named, recommended or cited for that exact question, which you can find out in about ten minutes.
This is the agency-specific version of a question every business now asks about itself: what is brand intelligence, the layer of AI-formed impressions that sits alongside your Google ranking and that most companies have never checked directly. For an agency it cuts two ways: your own visibility when a prospect asks who to hire, and your clients' visibility, which is the thing you were hired to improve.
Five shapes cover most of it: a city-specific search, a named alternative, a budget constraint, an industry fit and a platform fit. A prospect rarely types "good marketing agency." They type something closer to one of these, and each one pulls a different set of sources into the engine's answer.
| Question type | Example | What it tests |
|---|---|---|
| City-specific | "best SEO agency in Austin" | Local directories, reviews, your own location pages |
| Named alternative | "agencies like [competitor]" | Whether you appear next to the names buyers already know |
| Budget-constrained | "marketing agency for a $5k/month budget" | Whether your pricing or positioning is stated anywhere the engine can read |
| Industry-specific | "agency that specializes in B2B SaaS" | Case pages and content tied to that industry, not a generic services page |
| Platform-specific | "HubSpot agency partner near me" | Partner directories and certifications the engine can verify |
Why it matters: an agency that only checks "best marketing agency" is checking the one question shape it is least likely to win, and skipping the four it probably can.
Ask each of the five question types in ChatGPT, Claude and Perplexity with web search on, using your city, industry and services, then record named, recommended and cited for every answer. Named means your agency's name shows up anywhere in the text. Recommended means the engine puts you forward as a fit for that question, not a passing mention in a longer list. Cited means the engine links to one of your own pages as a source.
Do this: keep the record simple, a spreadsheet with one row per question and one column per engine is enough to see the pattern after two or three runs.
Mainly three things: a case or service page that answers the exact buyer question in its own words, a profile on the review site the engine already trusts for that category, and one consistent name and description across both. Engines that answer with web search open real pages at answer time. A services page that opens with company history rather than the answer to "agency for a $5k/month budget" gives the engine nothing to quote for that question, even if the agency itself fits.
For agencies specifically, the review sites that come up most are the ones built for this exact comparison, Clutch and G2-style directories among them. A profile left incomplete, or with an outdated service list, is one more place your agency's description can disagree with itself, which gives the engine less to agree on when it decides who to name.
Check it: the free external profiles check flags where your directory listings and your own site describe the agency differently.
Yes, and it doubles as proof of your own work. A client who never shows up when a buyer asks ChatGPT, Claude or Perplexity for a recommendation in their category is a gap the agency can point to directly, fix, and then show the client the before and after. Seeing which competitors AI recommends instead of a given site is the same method as the five-question check above, run against a client's category rather than "marketing agency."
Why it matters: an agency selling AI visibility work without having run this check on its own name first is a harder pitch to make.
None of this requires buying anything, and the whole check takes minutes:
Bottom line: most agencies have never asked AI the exact question a prospect would type, so the first check usually turns up something worth fixing either way.
The only way to know is to ask it yourself, in the exact words a buyer would use, with web search on, and read which agencies the answer actually names. A general sense that "we rank well" on Google says nothing about what ChatGPT, Claude or Perplexity say when someone asks for an agency.
Named means your agency's name appears anywhere in the answer. Recommended means the engine puts you forward as a good fit for the buyer's specific question, not just a mention in a list. Cited means the engine links to one of your pages as a source. An agency can be named without being recommended, and recommended without being cited.
Weekly is realistic for a small set of buyer questions, because the same question can return different names on different runs. A single check is a snapshot, not a measurement, so treat one good answer as encouraging rather than settled.
Yes, and for the same reason an agency sells SEO reporting: a client who is invisible to AI buyers is a problem the agency can point to and fix, which is also evidence of the agency's own work.
Open the sources the engine cites for that answer and read what they say about the competitor that your own site does not yet say about you clearly enough: a case page that answers the buyer's question, a profile on the review site the engine trusts, or a consistent entity across both.
The free check asks ChatGPT, Claude and Perplexity the buyer questions a prospect would type, and shows you who got named, recommended and cited. 3 minutes, evidence included.