AI SEO Agencies Are Selling the Wrong Kind of Visibility

Visibility in AI answers matters. But treating mentions as the product is a category mistake.

The newest bad habit in search is easy to spot.

An agency gets your brand mentioned in an AI answer, takes a screenshot, drops it into a deck, and calls it success. It looks futuristic. It sounds like page-one visibility for the AI era. But in many cases, it is closer to selling placement on Google’s second page: technically visible, occasionally flattering, and too often disconnected from the business outcome.

That does not mean AI search visibility is worthless. It means the market is confusing appearance with advantage.

AI search is no longer a fringe experiment. Google said in May 2025 that AI Overviews had expanded to more than 200 countries and territories and more than 40 languages. Google also says AI Mode breaks questions into subtopics and searches for each simultaneously, while Microsoft says Copilot Search uses Bing results plus additional searches issued on the user’s behalf to compile cited answers. This is a real discovery layer, and a growing one. But because it is a retrieval layer built from multiple sub-searches and supporting sources, the deliverable cannot simply be “we got you a mention.” It has to be “we made your brand retrievable, interpretable, and preferable on the prompts that matter.”

1. Mentions are the wrong KPI

The first problem is brutally simple: AI answer layers are built to reduce the need to click.

Pew Research found that Google users clicked a link inside an AI summary in just 1% of visits to pages that showed one, and users were more likely to end their browsing session entirely after seeing an AI summary than after seeing only traditional results. Seer Interactive’s 2025 study found that organic CTR for informational queries with AI Overviews fell 61% since mid-2024, while paid CTR on those same queries fell 68%. Even when brands were cited, the absolute numbers stayed low: Seer reported 0.70% organic CTR for queries where an AI Overview appeared and the brand was cited, versus 1.45% when no AI Overview appeared at all.

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. Seer found cited brands received 35% more organic clicks and 91% more paid clicks than brands not cited in the AI Overview. But that does not make mention counts a serious standalone KPI. It just 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.

To be fair, Google has said that links included in AI Overviews can get more clicks than if that same page had appeared as a traditional listing for the query, and Google also says clicks from search pages with AI Overviews are higher quality in the sense that users spend more time on site. Both things can be true at the link level. But strategy lives at the portfolio level, not the screenshot level. If the broader environment is compressing clicks, then “we got you mentioned” is still an incomplete answer to “did this move the business?”

Further Reading:

1. If you are comparing providers in this space, read Best AI Search Visibility Agencies, which explains how to evaluate GEO partners based on retrieval, citation-source analysis, and recommendation-stage performance rather than surface-level visibility claims.
2. For the source-layer behind stronger AI recommendations, read AI Citation Architecture Agency, which explains how citation gaps, third-party evidence, and public-source alignment shape whether a brand is trusted, cited, and recommended.
   

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 may use a “query fan-out” technique, issuing multiple related searches across subtopics and data sources, and that the set of responses and links can vary. Microsoft similarly says Copilot Search uses the user query plus additional searches to assemble a response. Research published on arXiv in 2025 found that AI search systems differ significantly 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.

The 2025 arXiv GEO study found that AI search shows a strong bias toward earned media and third-party authoritative sources over brand-owned and social content. The same paper 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.

In other words, 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 mentions won” report. And Seer’s own conversion-focused analysis makes the same point from a different 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.

Buy prompt coverage on the queries that shape category choice, vendor evaluation, objection handling, and shortlist formation.

Buy retrieval alignment so your pages are easy for machines to parse, compare, justify, and cite.

Buy corroboration across owned content and third-party sources, because AI search does not treat your website as the only authority on your brand.

Buy measurement tied to outcomes: 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.