1. Mentions are the wrong KPI
The first problem is brutally simple: AI answer layers are built to reduce the need to click.
Recent reporting and research point in the same direction. Pew Research found that Google users clicked links inside an AI summary in only a very small share of visits, and Seer Interactive reported sharp CTR declines for informational queries where AI Overviews appeared. Even when a brand was cited, the absolute click numbers remained low relative to environments without an AI Overview.
That is the part many agencies skip. They sell the screenshot and hide the denominator.
Yes, being cited is better than not being cited inside a suppressed environment. But that does not make mention counts a serious standalone KPI. It only means that if the answer layer is taking attention, you would rather be inside it than outside it. That is a very different claim from saying AI mentions are a business strategy.
Google has also said that links inside AI Overviews can receive more clicks than a traditional listing for the same query and that clicks from AI-enhanced search experiences can be higher quality. Those claims can be true at the link level. Strategy still has to live at the portfolio level. If the broader environment is compressing clicks, "we got you mentioned" is still an incomplete answer to "did this move the business?"
2. Mentions are unstable by design
A second problem is that AI mentions are not durable assets in the way many agencies pretend they are.
Google says AI Mode and AI Overviews can use a query fan-out technique, issuing multiple related searches across subtopics and data sources, and Microsoft says Copilot Search similarly performs additional searches to assemble a response. Research published in 2025 also found that AI search systems differ materially in freshness, domain diversity, cross-language stability, and sensitivity to phrasing.
In plain English, the mention you celebrate today may disappear tomorrow because the wording changed, the model changed, the supporting sources changed, or the engine interpreted the task differently.
That is why selling static mention counts in a dynamic retrieval environment is such a weak offer. One screenshot is not a moat. It is a moment.
3. Mentions do not equal authority
The third problem is deeper: mentions confuse presence with authority.
Recent GEO research found that AI search tends to favor earned media and third-party authoritative sources over brand-owned and social content. The same work argues that brands need machine scannability, justification, earned-media strength, and engine-specific strategy. That matters because AI systems are not just looking for a brand name to repeat. They are looking for enough supporting structure to justify inclusion.
So the real game is not sprinkling AI buzzwords across your service pages and hoping a model picks you up. The real game is building a footprint the system can trust: clear entities, explicit claims, original data, quotable frameworks, comparison pages, citations, third-party corroboration, and content architecture that makes your relevance legible to both humans and machines.
What matters is not whether your brand was mentioned once. What matters is whether the system has enough semantic evidence to keep retrieving you when the query changes shape.
4. There is no magic AI SEO switch
This is also why so much "AI SEO" advice sounds suspiciously like a repackaged gimmick economy.
Google's own documentation says there are no additional requirements to appear in AI Overviews or AI Mode, no special optimizations necessary beyond foundational SEO best practices, and no special schema or machine-readable file you need to create for inclusion. A page simply needs to be indexable, eligible for Search, and aligned with the same helpful, reliable, people-first principles Google already emphasizes.
So when an agency positions AI visibility as a mysterious new dark art, that should raise eyebrows. The opportunity is real, but the mechanics are less magical than the sales pitch suggests. Better information architecture, stronger entity clarity, clearer topical framing, structured proof, crawlable text, internal linking, and citation-ready content are still the foundation. The difference is that now they also shape how retrieval systems interpret and reuse your material.
5. Serious strategy starts where mention-selling ends
The final problem with mention-selling is measurement.
Google says traffic from AI features like AI Overviews and AI Mode is reported inside Search Console's overall Web search type. In other words, Google does not hand you a neat little AI report. And conversion-focused analysis from the market points to the same conclusion from another angle: visibility alone is not enough; you need revenue-impacting data to know when these surfaces actually matter.
That is why smart brands should stop buying mentions and start buying four harder things.
- Prompt coverage on the queries that shape category choice, vendor evaluation, objection handling, and shortlist formation.
- Retrieval alignment so your pages are easy for machines to parse, compare, justify, and cite.
- Corroboration across owned content and third-party sources, because AI search does not treat your website as the only authority on your brand.
- Measurement tied to outcomes, including assisted conversions, AI referral quality, branded search lift, pipeline influence, sales-call mention rate, and visibility on converting query classes.
That is the difference between AI theater and AI strategy.
The actual opportunity
Page-two rankings were never weak because nobody could technically find them. They were weak because they rarely changed the outcome.
AI mentions follow the same rule.
A mention can be real and still be commercially thin. A screenshot can look impressive and still sit miles away from revenue. And a brand can appear in AI answers without building any durable retrieval advantage at all.
The winners in AI search will not be the brands that buy the most screenshots. They will be the brands that reduce the distance between what they publish, what the web says about them, and what machine systems can confidently surface when a buyer asks the category a serious question.
That is the real work now.
Not mention selling. Not vanity reporting. Not synthetic page-one fantasies for the AI era.
The real work is retrieval alignment, citation readiness, earned authority, and closing the cosine gap between brand intent and machine interpretation.
That is a strategy.
The rest is just page two with better branding.