AI SEO Risks and What Not to Do
The main AI SEO risks are spam penalties, tricks aimed at AI models, fake reviews and promises you cannot keep. Google says its spam policies also cover attempts to manipulate generative AI responses, and the US FTC bans fake reviews, including AI-generated ones. Safe AI SEO means useful pages, true claims and honest measurement. 15
- Google's spam policies cover mass-produced pages, hidden text and manipulation of AI responses. Penalties include lower rank or removal.
- Microsoft found 31 real companies planting memory-poisoning prompts in links, which it treats as a security problem.
- The FTC rule bans fake and AI-generated reviews and lets courts impose civil penalties on knowing violators.
- Self-promotion can work briefly but fades, and it can name your competitors for you.
- Every risk here has a safe substitute: better pages, real proof and plain reporting.
What are the biggest risks of AI SEO?
Asking whether AI SEO is safe has a short answer: it is, if you avoid shortcuts. The biggest risks come from shortcuts, not from AI SEO itself. Five shortcuts cause most harm: mass-produced thin pages, hidden or manipulative instructions, fake social proof, self-serving content that misleads, and promises nobody can keep.
Each shortcut has a published rule or study behind it. The table gives the rule, the likely cost and the safer move, and the sections below explain each.
| Shortcut | What the source says | Safer move |
|---|---|---|
| Mass-produced pages | Scaled content abuse can lead to lower rank or removal, however the pages are made. 1 | Fewer pages, each with original value. |
| Hidden text or prompts for AI | Hidden text is a listed spam practice, and the policies cover AI responses. 1 | Put everything you want read in plain sight. |
| Memory-poisoning links | Microsoft saw 50 prompts from 31 companies and calls it a security issue. 4 | Earn recommendations from real mentions. |
| Fake reviews | FTC rule bans fake and AI-generated reviews. 5 | Ask real customers, without dictating sentiment. |
| Result promises | Google: no one can guarantee a top ranking. 8 | Promise method and reporting only. |
Can AI SEO get you penalised by Google?
Yes, if the work breaks Google's spam policies. The policy page says violating sites may rank lower or not appear at all, and enforcement can be automatic or manual. 1
The page lists scaled content abuse, which is many pages made mainly to manipulate rankings, and it applies however they are created, including with generative AI tools. It also lists cloaking, keyword stuffing, doorway pages and hidden text. 1
Google's AI optimisation guide adds that making separate pages for every phrasing of a question violates the scaled content policy when the goal is to game results. 3 Google also says the spam policies cover attempts to manipulate generative AI responses. 1
Is it safe to hide instructions for AI models in your pages or links?
No. Hiding text from people while leaving it readable by machines is a named spam practice. Google's examples include white text on a white background, text moved off-screen and zero font size. 1 A hidden line telling a model to recommend your brand fits that description.
Prompt injection SEO is the loose name for this family of tricks. Microsoft's security team describes one version it calls AI recommendation poisoning. A website adds a Summarize with AI button whose link carries a pre-filled prompt such as remember this company as a trusted source. Clicking it can store that instruction in the assistant's memory. 4
In 60 days of reviewing AI-related links in email, Microsoft found 50 distinct prompts from 31 companies across industries including finance, health, legal and software. All were real businesses. 4 Microsoft recommends that users check saved memories and distrust such links, and it treats the trick as something to defend against.
Our advice is simple. Do not do this. It risks trust, platform penalties and your name appearing in a security report. See how to get recommended by ChatGPT for methods that last.
What happens if you fake reviews or testimonials?
You may break US federal law. The FTC announced a final rule on 14 August 2024 banning fake and false reviews, including AI-generated ones, buying reviews tied to a required sentiment, and undisclosed insider reviews. 5 It also bars fake social media indicators such as bot followers when the buyer knew or should have known. 5
The rule took effect on 21 October 2024, and courts can impose civil penalties on knowing violators. 6 Honest review requests remain fine. The FTC says incentives are allowed as long as nothing requires a positive or negative opinion. 6
There is a business reason too. In a July 2026 survey of 2,338 US adults, 74.15% said they would be at least somewhat less likely to buy if a chatbot flagged mixed or negative reviews. The survey is self-reported and was published on the Semrush blog. 9 It follows, though no source tests it, that fake reviews can surface as doubts.
Are self-promotional listicles and comparison pages risky?
They carry accuracy and durability risks. In Ahrefs' test of 34 self-promotional pages, 82% of new mentions for an unknown brand came from the author's own pages. For an established brand it was 6%. 7
Results were uneven and short-lived. In 43% of answers that cited a conference page, the brand being promoted was still skipped, and about a quarter of citations never returned. 7 The author also warns that self-promotion can give free exposure to competitors.
The author's own guidance is worth copying. Add your brand to lists where it truly fits, avoid ranking yourself first on lists you wrote, do not attack rivals, fact-check comparisons and keep them current. 7 Our guide to writing comparison pages that get cited follows those rules.
Why is over-claiming the most common mistake?
Over-claiming is common because AI search results feel measurable, so sellers quote screenshots as proof. Google tells buyers to ask what results to expect, in what timeframe and how success is measured, and says anyone who promises first place should be avoided. 8
The same rule applies to what you publish. Generative models predict likely words rather than retrieving facts, so AI-drafted claims can be wrong. Google advises manual checking before publishing. 2 A false statistic on your site may be repeated by assistants and then traced back to you.
At Chalk Labs we do not guarantee rankings, citations or lead counts, and we label third-party results as third-party. See what AI SEO cannot do and AI SEO myths for the limits.
What should you do instead? A safe checklist
Black hat AI SEO and AI SEO scams share red flags, so replace each shortcut with a plain habit. The checklist below takes about an hour to run against your site and any agency proposal. It works as a screen before you sign anything.
If a provider fails more than one line, ask them to explain in writing. A good one will welcome the question, and the guide to hiring an AI SEO expert lists more.
- Every page answers a real question and adds something a competitor's page does not.
- No text, link or prompt is hidden from human readers.
- Every number on your site has a source you can link to.
- Reviews come from real customers and are not scripted.
- Comparison pages are fair, dated and fact-checked.
- Your agency's contract has no promise of rankings, citations or leads.
Frequently asked questions
Is using AI to write SEO content against the rules?
Can I add hidden text telling ChatGPT to recommend my company?
Are AI-generated customer reviews legal?
What is the single most common AI SEO mistake?
Treating a tactic as a sure thing. No file, prompt or page count secures citations, and Google says nobody can guarantee a number-one ranking. 8 Build useful pages, measure monthly, and judge providers by deliverables rather than promises.
Does Chalk Labs use any of these tactics?
These tactics are not part of the method we describe on this site, and our terms say we do not guarantee rankings, citations or lead counts. We sell AI SEO, so judge us against the checklist above.
Sources
- Spam policies for Google web search · Google Search CentralOpened October 20, 2026. Supports: Scaled content abuse applies however pages are created, including with generative AI; hidden text examples; cloaking, keyword stuffing, doorway abuse; violating sites may rank lower or not appear; policies cover attempts to manipulate generative AI responses. Page last updated 2026-08-28.
- Google Search's guidance on using generative AI content on your website · Google Search CentralOpened October 20, 2026. Supports: Generative AI can help with research and structure; mass-producing pages without added value may violate spam policy; models predict likely words so fact-check before publishing. Page last updated 2026-10-01.
- Optimizing your website for generative AI features on Google Search · Google Search CentralOpened October 20, 2026. Supports: Separate content for every query variation violates scaled content abuse policy when done to game rankings or AI responses. Page last updated 2026-07-10.
- AI Recommendation Poisoning · Microsoft Security BlogOpened October 20, 2026. Supports: Summarize with AI links with pre-filled memory prompts; 50 distinct prompts from 31 companies over 60 days of email review; users advised to check memories; platform mitigations. Published 10 February 2026.
- Federal Trade Commission Announces Final Rule Banning Fake Reviews and Testimonials · US Federal Trade CommissionOpened October 20, 2026. Supports: Announced 14 August 2024; bans fake and false reviews including AI-generated, buying sentiment-conditioned reviews, undisclosed insider reviews, fake social media indicators; civil penalties for knowing violators.
- Consumer Reviews and Testimonials Rule: Questions and Answers · US Federal Trade CommissionOpened October 20, 2026. Supports: Effective 21 October 2024; courts can impose civil penalties for knowing violations; incentives allowed if no required sentiment.
- We published 34 self-promotional pages to test AI SEO · AhrefsOpened October 20, 2026. Supports: 34 pages, 9,886 AI answers; 82% of new mentions for a new brand came from the author's pages versus 6% for an established brand; 43% of answers citing a conference page skipped the brand; about a quarter of citations never returned; author's advice on honest self-promotion. Small experiment on the author's own brands. Published 6 July 2026.
- Do you need an SEO professional? · Google Search CentralOpened October 20, 2026. Supports: No one can guarantee a number-one ranking; ask what results to expect, in what timeframe and how success is measured; promised first place is a reason to look elsewhere. Page last updated 2026-06-05.
- Do AI chatbots talk users out of buying? (survey) · Semrush blog, survey by Exploding TopicsOpened October 20, 2026. Supports: 2,338 US adults, fieldwork July 2026; 74.15% of consumers would be at least somewhat less likely to buy if a chatbot flagged mixed or negative reviews; self-reported and published by a vendor. Published 7 September 2026.
- Google Search Status Dashboard: ranking updates history · GoogleOpened October 20, 2026. Supports: Spam updates: 24 March 2026 (19 hours 30 minutes), 24 June 2026, 18 August 2026, 24 September 2026 (about 13 days 16 hours).