Buying mentions to rank in AI? Google now calls it spam

On May 15, 2026, Google updated the official page of its spam policies to clarify that they "apply to all of Google Search, including generative AI responses," per the Search Central documentation. The very definition of spam now includes a new line: "attempting to manipulate generative AI responses in Google Search." The same day, at Search Central Live in Shanghai, Gary Illyes (Google) warned head-on against buying or fabricating brand mentions designed to appear in AI responses.

The trigger is no accident. An SEO software platform had started selling automated "brand mention buying" with the explicit goal of placing those brands inside responses generated by Google, ChatGPT, Claude and others. Google's answer is a clear text and an intentional parallel: the bought mention is the new paid link.

What the text says, word for word

Google doesn't settle for a vague formula. The Search Central documentation is now explicit about fabricated mentions:

Operationally, this means no new rule was invented: Google simply extended its entire anti-spam arsenal to AI surfaces. Scaled content abuse (mass production of low-value pages, often AI-generated), site reputation abuse, link spam, cloaking and every other policy now govern what shows up in AI Overviews and AI Mode too. It's the direct continuation of the logic behind the May 2026 Core Update: one index, one quality bar, regardless of the surface.

Why this is a blow to an entire industry

For 18 months, a whole market has been built on the idea that "GEO" is a different game from SEO, with its own hacks. Vendors sell "mention packs," recommendation poisoning (stuffing third-party listicles to get cited), or brand injection into forums and niche sites. The bet: if LLMs read the web to build their answers, just flood the web with your brand to get cited.

Google dismantles that bet with the most familiar parallel an SEO can grasp: paid links. For twenty years, buying links has been detected, neutralized and penalized. The message is that a fabricated mention meets the same fate. This isn't a positive signal being neutralized; it's a behavior that can be penalized, up to demotion or removal from results.

What it concretely changes for an SMB

If you've signed with a vendor promising "X AI mentions per month" or "get cited by ChatGPT in 30 days via our network," now is the time to reread the contract. The risk is no longer just paying for an ineffective signal: it's paying for a signal that can now cost you your organic visibility.

1. Audit your existing mentions

List where your brand appears off your own site. Separate the earned (press, customer reviews, legitimate industry directories) from the fabricated (undisclosed sponsored posts, stuffed forums, bought listicles). If a large share of your citations comes from purchases, you have technical debt to clean up before it turns against you.

2. Bet on real citability, not volume

What gets a source cited by AI isn't how many times the brand appears: it's the quality and clarity of the information on your own domain. Direct answer at the top of the article, hard data, named sources, readable structure. That's the foundation of our AEO framework, and it's exactly what Google's documentation validates by implication.

3. Make your content machine-readable

Models cite what they understand quickly and unambiguously. A machine-readable information architecture (explicit headings, self-contained paragraphs, clean markup) carries more weight than ten mentions planted on third-party sites. And unlike bought mentions, this work belongs to you and can't be disavowed by Google.

4. Watch the sources that actually cite you

The Preferred Sources system Google is extending to AI Overviews rewards sources users deliberately choose, not those that self-promote. Build relationships with the media and communities in your sector so they cite you because you earn it, not because you paid for it.

What this update does not say

• No new detection system is announced. Google clarifies a rule; it doesn't unveil new technology. We don't know how precisely fabricated mentions are told apart from organic ones at scale. The operational uncertainty remains.
• The scope is Google. The rule targets AI Overviews and AI Mode. It says nothing about how ChatGPT, Claude or Perplexity handle fabricated mentions on their side. Those platforms have their own, sometimes opaque, logic.
• No list of "risky" vendors is published. Google names practices, not tools. It's on you to assess whether your agency's method counts as earned or fabricated.
• The penalty timeline isn't specified. As with paid links, the gap between "it's banned" and "it's penalized for you" can run into months. The absence of an immediate penalty is not a green light.

The Cicéro take

This update closes a debate that should never have opened. "GEO" was never a discipline parallel to SEO with its own shortcuts: it's the same fundamental work applied to new surfaces. Google just wrote it down in black and white. Vendors who sold mention volume will have to pivot or disappear; those who produced genuinely citable content have nothing to change.

That's exactly the methodology Cicero Studio has applied from the start: GEO audit + editorial production + automated semantic linking, with a minimum quality score of 90/100 on every piece, and zero fabricated mentions. Agency-quality work, software-grade productivity, from €250 to €1,800 per month. If you want to know whether your AI visibility rests on clean foundations or on risky signals, our complete GEO audit method is documented here.

Alexis Dollé

CEO & Founder

Growth and SEO content strategist, I founded Cicéro to help businesses build lasting organic visibility, on Google and in AI-generated answers alike. Every piece of content we produce is designed to convert, not just to exist.

LinkedIn