Google can count the link, not the conclusion
Search Console can now show when a site appears inside Google’s generative AI features. For reputation teams, that is useful but narrow. The report can show that a company page was exposed. It does not show whether the answer helped the company, criticized it or used the page as background for someone else’s recommendation.
Exposure is not the same as representation
A link to a corporate page can support a favorable description, substantiate criticism or simply supply factual context. Each case can create the same impression in the report, while the user carries away a different judgment.
That is why the new metric belongs inside AI reputation work, not ordinary search reporting alone. The question is not only whether Google surfaced the page. It is what the answer made the page mean.
What this piece covers
- Why generative AI impressions measure source exposure rather than the reputational conclusion shown to the user.
- How a crisis, lawsuit, recall or executive dispute can increase AI impressions while weakening trust.
- Why owned-page exposure can reveal where Google is looking for evidence, without proving that the answer favors the company.
- How reputation teams should combine Search Console data with sampled answer review across high-value decision queries.
A crisis can make the number look better
Corporate pages often receive more attention during controversy. An incident statement, regulatory notice or executive response can become more useful to Google precisely because users are asking about a problem. The Search Console metric can rise while the company’s reputation is under pressure.
That is the same measurement trap behind AI summaries that shape perception before users inspect the source and reputation formed before the click. The page appeared, but the company still does not know what judgment surrounded it.
The page view can help map demand for evidence
The most useful part of the report may be the page dimension. If Google repeatedly surfaces product documentation, policy pages, executive biographies or old corporate statements inside AI experiences, the company can see which parts of its public record are being pulled into generative search.
That does not make the aggregate count a reputation score. It makes the report a source-demand map. The same issue appears when the citation matters more than the search position and when third-party evidence supplies the stronger material for an AI answer.
Recommendation needs a separate test
Search Console cannot tell whether the company was recommended, merely mentioned, compared unfavorably or excluded from the final suggestion. A brand can receive an impression while another company receives the useful conclusion.
Reputation teams therefore need a representation review above the Search Console metric: controlled prompts, archived answer text and comparison across the AI surfaces that matter to customers, investors, partners and candidates. The same standard applies when teams test whether AI agents select the company consistently.
Without that second record, companies can mistake exposure for endorsement. The risk is higher because AI outputs already influence high-stakes decisions, and because a company can be described incorrectly even when public material exists.
