Search Console

You have AI impressions in Search Console. Does AI recommend you, though?

No, not on its own. An AI impression in Search Console means Google pulled a snippet from your page into an AI Overview; it says nothing about whether ChatGPT, Claude or Perplexity ever name you as the recommendation, a separate check that takes about three minutes.

By Reflexa Technologies — the team building Reflexa, the AI Visibility Platform · September 15, 2026

In brief

What does an AI impression in Search Console actually show?

It shows that Google's system pulled a snippet from your page into one of its own AI features, nothing more. Google finished rolling out the new AI performance report to every Search Console property worldwide on August 31, 2026, after announcing it in June 2026. It groups impressions from AI Overviews, AI Mode and the generative AI features in Discover into one count, broken down by page, country and device, with no query-level data and no clicks. We covered the report itself in detail in Google Search Console's new AI performance report.

Proof: Google Search Central blog, "Introducing Search generative AI performance reports in Search Console" (June 2026); Search Engine Land, "Google Search Console AI performance reports and Search generative AI control rolling out globally" (August 2026).

Do this: read a rising impression count as "Google considered this page," not as "AI recommends this business." The two are measuring different things entirely.

What does it actually mean for AI to recommend you?

A recommendation is the one outcome that puts you in front of the buyer, and it's a different event from being cited or mentioned. Reflexa tracks three separate outcomes for this reason: a Mention (your brand is named somewhere in the answer), a Citation (your page is listed among the sources AI drew on), and a Recommendation (your brand is the one AI actually suggests the buyer choose). An answer can cite your page as a source while recommending a competitor by name, or mention you in passing without ever suggesting a buyer pick you. None of that shows up in a Search Console impression count, which only tracks Google's own AI surfaces and can't see ChatGPT, Claude or Perplexity at all, part of the wider picture in AI search statistics 2026.

Bottom line: an impression, a mention, a citation and a recommendation are four different counts. Only the last one is the buyer's answer.

How big is the gap between being cited and being recommended?

Large enough that even a category leader can be cited far less often than it's mentioned or recommended. In the September 2026 project-management edition of the Reflexa AI Visibility Index (100 fixed buyer questions, asked three times each to ChatGPT, Claude and Perplexity), Asana was the most recommended tool in the category, yet the three counts told three different stories.

SignalAsana's rateRaw count
Mentioned somewhere in the answer51%452 of 882
Recommended in the top 341%359 of 882
Own site cited as a source18%158 of 882

Proof: Reflexa AI Visibility Index, September 2026, project-management edition.

Rule of thumb: mentions, citations and recommendations move independently, so a strong number on one of them tells you nothing about the other two.

How do I check whether ChatGPT, Claude or Perplexity actually recommend me?

Ask them the same buying question a real customer would, in a fresh chat, and read what comes back. Open ChatGPT, Claude and Perplexity separately (logged out or in a new conversation) and ask 3-5 variations of "what's the best [your category] for [your typical buyer]?" or "[your category] alternatives to [a known competitor]." Write down whether you're named, not named, or named but wrong, and note which sources each engine cites when it skips you. Doing this by hand across three engines takes longer than doing it once; the free check at dashboard.reflexatechnologies.com/audit runs the same kind of buyer questions across all three at once and returns the answer text and sources in about three minutes.

Shortcut: run each question at least three times before drawing a conclusion. Engines are non-deterministic, so one lucky answer is a snapshot, not a measurement. See how often to re-check.

What to do this week

None of the steps below need a purchase, and most take a few minutes each:

Do this: read the two reports together, never one alone: Search Console tells you Google's own AI pulled from your site, the free check tells you whether any engine actually recommended you once it did.

Sources: Google Search Central blog, generative AI performance reports (June 2026) · Search Engine Land, AI performance reports and generative AI control rolling out globally (August 2026) · Reflexa AI Visibility Index, September 2026

Keep reading

Frequently asked questions

Does an AI impression in Search Console mean ChatGPT recommends me too?

No. Search Console's AI performance report only covers Google's own AI Overviews and AI Mode. It has no data on ChatGPT, Claude or Perplexity, so an impression there says nothing about whether those engines ever name you.

What's the difference between a mention, a citation and a recommendation?

A mention means your brand is named somewhere in the answer. A citation means your page is listed as a source AI drew on. A recommendation means AI actually suggests the buyer choose you. A page can be cited without being recommended, and a brand can be mentioned without being the pick.

How big is the gap between being cited and being recommended, in practice?

In Reflexa's AI Visibility Index for project management (September 2026), the category leader was mentioned in 51% of eligible answers (452 of 882) but had its own site cited in only 18% (158 of 882) and reached a top-3 recommendation in 41% (359 of 882). All three numbers move independently.

How do I check whether ChatGPT, Claude or Perplexity actually recommend me?

Ask each engine 3-5 real buyer questions yourself in a fresh chat, or run the free check at dashboard.reflexatechnologies.com/audit, which does it across ChatGPT, Claude and Perplexity in about three minutes and shows the full answer text and sources.

How often should I re-check, instead of trusting one answer?

Weekly is a reasonable baseline. AI answers are non-deterministic and shift as web results change, so a single run on a single engine is a snapshot, not a measurement.

Search Console shows an impression. See the actual answer.

The free check runs real buyer questions across ChatGPT, Claude and Perplexity — 3 minutes, evidence included.

Run the free check →