AI Engines Named Us Almost Only When They Had Read Our Page: A 2026 Study of 2,895 AI Answers

How ChatGPT decides which brands to recommend

Most advice on this question stops at a bit of training data and a bit of live search. Both play a part. In our data, almost every answer that named LoudFace had also read one of our pages.

The engines tell you how they answer. OpenAI says "ChatGPT can search the web to answer questions with current information and links to relevant sources," and that it may search automatically "when your question would benefit from current information." When it uses a search partner, it "typically rewrites your query into one or more targeted queries" and sends those out. Perplexity says it "uses advanced AI to search the internet in real-time." Google says AI Overviews and AI Mode may issue "multiple related searches across subtopics and data sources", and that a page has to be indexed and eligible for Search to appear as a supporting link.

So for a buyer question like "which AEO agency should a fintech hire?", the engine runs a few searches, reads a handful of pages, and writes an answer from what it read. A brand can still be named in an answer with no sources recorded, and our data shows that a little. For LoudFace, though, being named and being read went together in almost every answer.

We read eight ranking pages across this question and its variants on 24 September 2026. None of them measured it. So we did.

What we measured

We track how AI engines answer our buyers' questions in Peec. For one week, 17 to 23 September 2026, we took every tracked answer on ChatGPT, Perplexity and Google AI Overviews and asked two questions of each one:

1. Did it name LoudFace? LoudFace appears among the brands in the answer.
2. Had it read LoudFace? A loudface.co page appears among the sources the answer used.

We left out the four tracked prompts that contain the word LoudFace, because an answer to "Is LoudFace a good AEO agency for B2B SaaS?" names us by definition. That left 3,051 answers. We could read 2,895 of them in full. The other 156 could not be retrieved from the tracker and are excluded rather than guessed.

The result: 624 of 631 namings were in answers that read our site

LoudFace was named in 631 of the 2,895 answers. In 624 of them, the engine had read at least one loudface.co page. That is 98.9%.

Turn it around and it is starker. 1,819 answers read no LoudFace page. Seven of them named us, 0.4%. For practical purposes, when the engine did not open one of our pages, we were almost never in the answer.

This is the second time we have seen it. An earlier read of 120 answers, from 26 August to 1 September 2026, found 17 namings, all in answers that had read our site. The week-long read covers 24 times as many answers and lands in the same place.

Engine by engine

Google AI Overviews named us 138 times, and every one of those answers had read our site. ChatGPT read our pages most often, in 650 answers, and named us in 58.9% of them. Perplexity read us in 248 answers and named us in 41.5%. It is the engine where a read most often ends without a mention.

The seven exceptions

We opened all seven answers that named LoudFace without a loudface.co page in their sources.

• Four were ChatGPT answers with no sources recorded at all. They answered questions about fintech infrastructure agencies, AI Mode visibility and a monthly budget range. These may be answers from memory, or answers whose sources the tracker did not capture. Four looks like noise next to all 1,819 answers that read no LoudFace page. The fair comparison is ChatGPT's 16 answers with no sources at all, and 4 of them named us. Naming with no sources recorded is small, but it is not zero.
• Three were Perplexity answers that named us from other people's lists, such as a roundup of cybersecurity SEO agencies. Other people's pages can put us in an answer. In this week, that happened three times.

In our August read we saw no third-party namings at all, so three is a small change.

Being read went with almost every naming. It is not enough.

1,076 answers read a loudface.co page. 452 of them still did not name us. Reading gets you into the running. The page still has to win.

The pages with the highest share of namings were mostly the ones about LoudFace itself:

Most answers that read a page saying clearly who we are and who we serve also named us. A broad list with many agencies on it got read more, and under half of those answers named us. One answer can read several pages, so the rows overlap.

This is an observation, not a controlled test. The service pages may get read mostly on questions where we are already a strong fit. But it matches what we have seen engines lift about LoudFace before: a specific positioning sentence, a published number, a named methodology.

How to get ChatGPT to recommend your business

If being named goes with being read, most of the usual advice gets simpler. Here is the order we would work in.

1. Find the questions your buyers ask an engine. Not your keywords. The actual questions: "best payroll software for a growing startup", "who runs security audits for SaaS companies". Those are what the engine searches for.
2. Make sure a page of yours answers each one directly. One page per question cluster, with the answer in the first screen. If no page of yours answers the question, brand strength alone rarely gets you read for it.
3. Get that page indexed and crawlable. Google says a page must be indexed and eligible for Search to be a supporting link. ChatGPT's search sometimes relies on partner search providers. A page the engines cannot fetch is a page they cannot read.
4. Say who you are on the page, in one plain sentence. The pages that most often went with a naming for us state what we do and for whom. Make your name the obvious answer on the page the engine reads.
5. Then work on other people's lists. Three answers in a week named us with only other people's pages in their sources. Work that lever once your own pages are in place.
6. Measure namings per engine, on the same questions, every week. A single check tells you very little. The split by engine is where the work is: our read-to-naming rate runs from 41.5% on Perplexity to 77.5% on Google AI Overviews.

If you are not sure your brand is even recognised correctly, start with entity disambiguation. If you want the full step-by-step for one engine, our ChatGPT playbook for B2B SaaS covers indexing, crawlers and page structure in detail.

Why ChatGPT doesn't mention your brand

Read the result backwards and the usual reasons line up.

• No page of yours answers the question. The engine searches the question, finds other companies' pages, and names them. In our data this is the usual pattern, and it is the easiest to fix.
• Your page gets found but does not say who you are. Our lists got read 196 times and named us in 84. A page where you are one of twenty entries does less for you than a page about you.
• The engine cannot fetch your page. It is missing from the index, or the crawlers are blocked.
• You only appear on other people's pages. That did work three times in a week for us. It is the slow route.

None of these is fixed by being more famous in the abstract. They are fixed page by page.

What this changes about measuring AI visibility

Visibility, the share of answers that name you, is still the number that matters. Peec's definition is simple: "If your brand is mentioned in 4 of 10 chats, your visibility is 40%." What this study adds is the number underneath it. Track how often the engines read your pages, and how often a read becomes a naming. The first tells you whether you are in the running for a question. The second tells you whether your page makes the case.

That is how we read our own program. Over a longer run, our own case study tracks LoudFace's non-branded AI visibility from 0.13% in April 2026 to 21.5% over 18 to 28 September 2026 (ChatGPT 38.8%, Google AI Overviews 14.3%, Perplexity 11.2%), measured after a 17 September prompt cut that lifts the reading and over a different window from this study. For a wider view of how buying behaviour is shifting from rankings to recommendations, see the new search funnel.

If you want to know which of your pages the engines read for your buyers' questions, and how often that turns into a naming, start with an AI visibility audit or talk to us about a GEO program.