Google’s Search team published guidance on 2026-05-30 warning brands against “inauthentic mentions” as a tactic for AI search visibility. The summary, plain language: buying placements, manufactured citations, paid “thought leadership” on low-trust sites, and any pattern of mentions that does not match real-world brand discussion will get flagged. Inauthentic mentions may look useful in the short term. They create long-term SEO risk.
This matters today because for the last 60 days, every Shopify operator has been pitched the same offer by GEO services agencies: pay us 3 to 8 thousand dollars a month and we will buy you into ChatGPT, Perplexity, and Gemini citations. The methodology is rarely transparent. The placements often are exactly what Google just warned about.
What Google’s guidance actually says
Google’s framing is that AI search engines, including Google’s own AI Overviews, build a brand authority profile from the pattern of mentions across the open web. That profile flags when the pattern looks engineered: bursts of new mentions on low-traffic blogs, identical language across unrelated sites, citations from properties that have no editorial relationship with the brand. The AI search engines deweight or remove these brands from candidate sets. Google warns that the deweighting carries over to standard search ranking too.
For a Shopify operator, this means a paid GEO campaign is not just wasted spend. It is potentially negative spend.
Three authentic alternatives that build the same citation signal
The AI search engines are not against brand mentions. They are against the manufactured kind. Brands that earn AI citations all do versions of three things.
1. FAQ schema seeded with the actual questions customers ask
Every Shopify store has a customer support inbox or a chat log. That inbox is the highest-quality source of real customer questions in the world for that brand’s category. Take the 20 most-asked questions, write a 3 to 6 sentence answer to each, mark up as FAQPage schema, and put it on the relevant product pages.
This is the signal AI search engines parse most reliably for product-recommendation prompts. ChatGPT and Perplexity quote FAQ-schema’d answers verbatim. Most stores either skip FAQ schema or fill it with marketing copy. Real customer questions, written in real customer language, is the cheat code.
2. A consistent expert-quote flow
Brands that get cited on Grok and Claude almost always have a regular pattern of expert quotes embedded in their content. This does not require a PR budget. The expert can be the brand’s own formulator, founder, head of product, or a long-term advisor. The format is: one named person, one specific claim, one verifiable credential. Embedded as quote markup with a clear attribution.
Over 90 days, even one quote per product page or blog post builds the signal that AI search engines use to grade citation-worthiness. The competing approach (buying mentions on third-party sites) takes the same 90 days and creates the risk Google just warned about.
3. Off-site presence on the surfaces AI search engines actually crawl
The brands that win Perplexity and ChatGPT citations are present on YouTube (demos, even short ones), review aggregators (Trustpilot, homewisereview.com, vertical-specific review sites), and at least one community forum where their category is discussed (vertical subreddits, Reddit, Shopify Community, etc).
Posting on these surfaces is engagement, not a transaction. It is slower than buying mentions. It survives Google’s policy because it is what the policy was trying to protect.
How to measure if it is working
The trap with all three of these is that they take 3 to 12 weeks to compound, and there is no built-in feedback loop. Most operators give up at week 4 because there is no visible change.
The fix is to measure citation rate directly. Pick 10 to 20 high-intent prompts in your category (the kind a buyer with a credit card would type), run them across ChatGPT, Gemini, Perplexity, Claude, and Grok every week, and track which brands get cited. When your FAQ schema starts showing up in ChatGPT answers, or your expert quote starts getting attributed in Claude, you will see it in the citation rate before you see it in revenue.
Citelix measures exactly this. We scan your store across the 5 AI platforms every day and tell you which prompts your competitors win and what changed. The free scan at citelix.com/scan runs your first prompt by hand within 24 hours, no card required.
The TL;DR
Do not buy mentions. Google flagged this on 2026-05-30 as a long-term penalty risk. Build the same citation signal authentically through FAQ schema (from real customer questions), expert quotes (from real internal experts), and presence on the off-site surfaces AI engines actually crawl. Measure the citation rate weekly so you can see it compound.
The brands that win AI search in the next 12 months are the ones that ignore the GEO services pitch and build the structured signal themselves.