Why: Sources, Search, Memory and Engine Briefs

A gap in Answer Share has causes, and they call for different changes. Why looks at every topic and every competitor through three lenses, and sets beside them what each engine says about you. Each lens rests on evidence one click away: the answers, the pages they cited, and the searches the engines ran.

The reasons

At the top of a topic’s Why, the lenses are summed up as plain sentences, strongest first, for example:

• “Of the pages engines read for this topic, 14 of 31 name Fabrikam; 2 name you.”
• “When engines search for this topic they run 6 searches; Fabrikam is in Google’s top 10 for 5 of them, you for none.”
• “Without searching, engines name Fabrikam in 70% of answers and you in 0%.”

(The figures are illustrative.) Each reason is computed from the numbers, never written by a model, and marked strong, some or weak. A weak reason says what is missing, for example that only a few of the cited pages could be read. Each topic row on the Overview carries its strongest reason.

Sources: the pages engines read

For each topic, the third-party pages the engines cited or read in their answers, most-read first. Every page is read for the brands it names, in order, and whether it is a ranked list.

• Pages: each with its kind (a list, a review, a comparison, an article, a forum thread, a directory, a vendor page), the answers and engines that used it, the brands it names, whether it names you and the leader, and when it last changed.
• Gaps: pages read for the topic that name a competitor and not you. A page you could not be named on is never a gap: a competitor’s own page (you cannot be listed on Fabrikam’s own site), an app’s store listing, a company’s product page, documentation, a forum thread, an encyclopedia, or a “best of” list its own publisher is one of the options on. Each page also carries its language, and one in another language than your market’s is marked.
• Coverage: on how many of the read pages each brand appears, and how many answers cite those pages.

Being on the pages engines read is the strongest lever we have measured: in our research, brands named on many of the pages the engines read for a question were named in answers far more often than brands named on none. That is why most off-site plays start here.

Pages are read politely, with robots.txt honoured. Pages behind a login, and platforms that do not allow it (forums and social networks among them), are counted as cited with their brand list unknown. A page’s versions are kept, which is how a page newly naming a brand is dated. Your own pages are left out of Sources; they are the site audit’s.

Sources, under Competitors, lists the pages to be on across all your topics, gaps first.

Search: the searches engines run

To answer, most engines search the web, and several report the searches they ran. Those searches are rarely your question word for word: an engine turns a question into the searches it expects to find good pages for, very often “best …” lists. For each topic, the most frequent of those searches are looked up in the index the engine reads:

Index	Stands for
Google	Gemini, Google AI Overviews and AI Mode
Bing	Copilot Search, and in part ChatGPT
Brave Search	Claude

For each search: where you and the leader rank in the top 10 of each index, and the ten results. An engine whose index is not checked is shown as “not checked”. A results page whose results have little to do with the search is treated as not checked rather than as a page you are absent from.

A brand counts as present when one of its own pages is a result, or when a result’s title names it (a third-party page about it). Brands the engines recommend rank for the engines’ own searches far more often than brands they pass over.

Memory: what engines know without searching

A small share of the poll’s questions is asked with search off, of the engines that can answer that way. Those answers show what each engine knows from training alone, and which brands it names when it does not look anything up. They never count toward Answer Share.

Memory and search are different populations: the brands an engine names from memory overlap only partly with the ones it names after searching. A gap in memory moves slowly and is closed by presence across the web over time; a gap in search can close in days. When memory confuses your brand with another company, Vercatus says so, and an entity play follows.

Engine Briefs: what the engine says

Once a week, each engine is asked directly about you, with search on and, where it can, with search off:

1. What do you know about the brand?
2. Would you recommend it for its main topic, and why or why not?
3. Who are the best options for that topic, and what would make the brand the first choice?
4. What are its weaknesses, or the reasons people choose a competitor instead?

Briefs, under Competitors, shows one card per engine:

• whether it knows you well, partly or not at all, and whether it would recommend you;
• one paragraph, the engine’s view in its own terms;
• the facts it states, and the reasons it gives for and against you;
• what it says would change its mind, with a mark where the evidence agrees;
• what changed since the previous brief;
• Show the probes: the full answers behind the summary.

Every card says: “What the engine says. Engines’ own explanations often differ from what moves their answers; the lenses show the evidence.” That label matters. When engines are asked what would make them recommend a brand, they mostly name the product and the price; the data shows that presence on the pages they read moves their answers more. So a brief is perception. Its claims become plays only when the evidence agrees, and a fact an engine gets wrong becomes a fact correction play.

Competitors

Competitors lists the brands your topics’ answers pick instead of you: each with its Answer Share and change, the topics where it leads you, how many read pages name it and how many of those leave you out, how many checked searches it ranks in, and its share without search. A competitor’s page sets it beside you per topic, with the sources that name it, its searches, its memory share, what each engine says about it, and what changed recently: pages that newly name or dropped it, clear share moves, and graded natural experiments.

See Competitors, Sources and Briefs for working with these screens, and Plays for what the evidence turns into.