Don't try to trick AI answers: hidden prompts, fake reviews and why they backfire
By the SEOCells team, OmCells R&D Labs LLP · Published
Tricks aimed at AI answers, such as hidden instructions, invisible text, fake system messages, manufactured consensus, fake reviews, false urgency and sabotaging competitors, tend to backfire. Engines work to detect them, the penalty can hit your whole brand, and fake reviews are illegal in several countries. Visible, honest content is the safer path.
Key points
- Text addressed to AI that people cannot see is a recognised manipulation pattern, and engines look for it.
- Search spam policies already cover hidden text, cloaking and link schemes.
- Fake reviews and testimonials break consumer-protection law in several countries.
- A penalty may not stop at one page; trust in the whole brand can suffer.
- Check your own pages too: hidden instructions can arrive through plug-ins, widgets or user content.
What kind of tricks are we talking about?
As AI answers became a source of customers, an old idea came back in new clothes: if you cannot earn the recommendation, try to force it. We see the same handful of tactics come up again and again.
| Trick | What it looks like | Why it backfires |
|---|---|---|
| Hidden instructions | White-on-white text or hidden elements telling the AI to recommend you | A known attack pattern; engines filter it and may distrust the site |
| Fake system messages | Text pretending to be an instruction from the AI company | Same as above, and obvious to anyone who looks |
| Invisible keyword or praise text | Text people cannot see, written for machines | Hidden text breaks search spam policies |
| Synthetic consensus | Many look-alike sites or posts all praising the same business | Link schemes and scaled spam; patterns are detectable |
| Fake reviews | Bought, invented or incentivised reviews | Illegal in several countries; platforms remove them |
| False urgency or scarcity | “Only 2 left!” that is always true | Misleading claims; erodes trust with people and engines |
| Competitor sabotage | Planting false claims or fake negative reviews about rivals | Legal exposure and reputational damage if traced |
Do these tricks work?
Researchers have shown that some of them can influence AI answers in experiments and, at times, on live systems. That is exactly why engines treat them as attacks and invest in detecting them. A trick that works today is a pattern that gets filtered tomorrow, and the site that used it has left evidence behind.
There is also a simpler problem. If everyone hides instructions on their pages, the answers get worse for everyone, and the engines have every reason to come down hard. Betting your business on a loophole the platform is actively closing is a poor bet.
What is the real risk?
- Detection: engines and search spam teams look for hidden text, cloaking and manipulation patterns.
- Brand-wide suppression: trust is often judged at the level of the site or the business, not just the page.
- Law: in the United States, the FTC's 2024 rule bans fake reviews and testimonials, and other countries have similar consumer-protection rules.
- Platform action: review sites and marketplaces remove fake reviews and can flag the business.
- People: customers who notice manipulation tell others, and that story spreads faster than any citation.
Could my site contain hidden instructions without me knowing?
Yes, and this is the part that surprises owners. Hidden text can arrive through a theme or plug-in from an unreliable source, a third-party widget, an agency that tried a shortcut years ago, or comments and reviews posted on your own pages. We have seen pages carrying text the owner had never read.
The SEOCells AEO Report includes a check for hidden instructions on your page and flags any it finds, so you can remove them before they cause trouble. We do not publish how the check works, partly so it is harder to evade.
How can you check your own pages?
- View the page source and search for words addressed to machines: “AI”, “assistant”, “ignore”, “instructions”, “recommend”.
- Look for text styled to be invisible: same colour as the background, tiny font sizes, positioned off-screen.
- Review plug-ins and widgets you no longer use, and remove them.
- Moderate comments and user reviews on your site.
- Check that your structured data says the same as the visible page: names, prices, ratings, dates.
What if an agency offers to do this for me?
Ask them exactly what they will put on your site and where. If the answer includes text visitors will not see, reviews they will “arrange”, or networks of sites that will mention you, walk away. Your name is on the domain, not theirs, and you will carry the consequences long after the contract ends.
What works instead?
The unglamorous things. Readable pages with clear answers. Real reviews from real customers, asked for fairly. Coverage on sites that chose to write about you. Claims a reader can check. None of it is quick, but none of it can be taken away by the next spam update either.
Questions answered
Is it against the rules to hide text for AI on my website?
Hidden text intended to manipulate search systems breaks Google's spam policies, and AI engines treat hidden instructions as an attack. The risk is demotion or loss of trust for the whole site.
Are fake reviews illegal?
In several countries, yes. In the United States the FTC's 2024 rule bans fake reviews and testimonials, and other jurisdictions have consumer-protection laws against them.
Could a plug-in add hidden instructions to my pages?
It can happen through untrusted themes, plug-ins, widgets or user-generated content. Review what is installed and check your page source.
Does SEOCells check for hidden instructions?
Yes. The AEO Report checks your page for hidden instructions aimed at AI and flags any it finds.
What if a competitor posts fake reviews about me?
Report them to the platform, keep records, and seek legal advice if needed. Respond publicly and calmly to set the record straight.
Sources
- Google Search Central — Spam policies for Google web search
- U.S. Code of Federal Regulations, 16 CFR Part 465 — Use of Consumer Reviews and Testimonials (the FTC's fake-reviews rule)
- Google Search Central — Creating helpful, reliable, people-first content
- Aggarwal et al. (2024), the GEO paper, KDD 2024 (arXiv:2311.09735)
Search engines and AI assistants change often; check the linked documentation for the latest details. Nothing in this post is a promise of rankings, citations or results.
Check your page for hidden instructions before they cause trouble.
See the AEO Report