Can You Trick ChatGPT Into Recommending Your Business?

No, not in any way that lasts. The known tricks are prompt injection, fabricated statistics, and manufactured reviews, and each can nudge an answer briefly before the source gets discounted. Money does not work either. OpenAI began testing ads in ChatGPT in February 2026. The company states the ads "do not influence the answers ChatGPT gives you." The durable way in is a site the engines can read, quote, and verify.

What tricks do people actually try?

Four tricks dominate the pitches and forum threads: prompt injection, fabricated statistics, manufactured reviews and awards, and flooding the web with machine-written pages.

The trick	The promise	What actually happens
Prompt injection (hidden text instructing the model)	"The AI will follow your instruction"	Engines increasingly separate commands from content
Fabricated statistics	"Engines love quoting numbers"	Uncorroborated numbers flag the domain as a risk
Manufactured reviews and awards	"Authority signals win recommendations"	Unverifiable signals get cross-checked and discounted
Flooding with machine-written pages	"More pages, more chances"	Reads as scaled spam to Google and engines alike

Each tactic is dissected properly in black hat GEO, and Google now treats the flooding pattern as scaled content abuse under its spam policies. The short version is the same for all four. They manufacture a signal instead of earning one, and the manufacturing is what gets detected.

Can you pay to become the recommendation?

No. There is no price list for a ChatGPT recommendation, from OpenAI or from anyone else.

The ad test OpenAI announced in February 2026 places ads at the bottom of answers, for logged-in users in the United States. The same announcement is explicit that those ads do not change what the model says or recommends. Sponsorship buys placement beside the answer, never a place inside it.

Anyone selling "guaranteed AI mentions" deserves the same skepticism as anyone selling a guaranteed number-one ranking on Google. Neither seller controls the system they are promising to control.

Why do the tricks stop working?

The tricks stop working because answer engines score trust at the domain level and check claims against independent sources. A statistic that exists nowhere else, or a review pattern nothing corroborates, makes your page the riskier quote rather than the easier one. The full selection mechanics are in how AI engines choose what to cite.

There is a second, quieter problem. Every answer is generated fresh, so a nudge that worked in one session holds no position anywhere. You have rented a coin flip, and the discount that follows detection applies to every future flip.

Detection also ends differently here than in search. A Google penalty comes with a report and a review process. A model that stops trusting your domain files nothing and tells nobody.

What actually changes what ChatGPT recommends?

The honest levers are the four gaps covered in why ChatGPT recommends your competitor. A site its crawlers can read. Pages that answer real buyer questions. A clear statement of who and where you are, and mentions on sites the engine also reads.

None of that is a trick, which is exactly why it holds. An engine that re-checks your pages finds the same verifiable substance every time, so the answer-engine work compounds instead of expiring.

A pattern we've noticed: clients hope visibility is something you buy once, and the trick-based pitch is priced to that hope. The businesses that actually get recommended treat it as publishing work inside a long-term SEO plan, done steadily by the playbook that earns AI mentions.

Frequently asked questions

Does hidden text on a website really influence ChatGPT?

Occasionally, briefly, and at real cost. Some injected instructions have slipped through in the past, which keeps the tactic alive in sales pitches. Engines now work to strip instructions from retrieved content. A domain caught carrying them risks being discounted across every query, a poor trade for a trick that cannot hold a position.

Can I buy an ad so ChatGPT recommends me?

No. OpenAI's ad test keeps ads visually separate from the answer, and the company states the ads do not influence what ChatGPT says. An ad can put your name on the screen next to an answer that recommends someone else. Only the retrieval layer decides whose pages get quoted.

My competitor seems to be gaming it. Should I copy them?

No, for two reasons. First, what looks like gaming is usually readability: their site answers questions machines can quote, and yours may not. Second, if they genuinely are manufacturing signals, the discount lands on their domain sooner or later, and copying them buys you the same exposure without the head start.

How would I know if my domain has been discounted?

You would not get a notice, which is the uncomfortable part. The only practical signal is the trend line: ask the engines about your business and your category monthly, record the answers, and watch for a slide. That baseline check is exactly what our free AI Visibility Report documents, and you can run a manual version yourself today.

How long does the honest route take?

Months, on roughly the same horizon as ordinary SEO's six-month minimum. Crawling, indexing, and third-party corroboration each take their own time. Citations on one specific, well-answered question tend to arrive earlier than broad recommendations, so the work pays out in stages rather than all at once.

Earn the answer instead

Storming Solutions runs SEO, AEO, and GEO for Malaysian businesses from Kuala Lumpur, and we turn down work that asks us to manufacture signals. Our reasoning is commercial, not pious: a recommendation an engine was tricked into is rented, and the engine sets the eviction date.

Want to know what ChatGPT already says about your business before you spend anything? Ask for the free AI Visibility Report on our contact page. The earning version of the work is scoped on our SEO, AEO, and GEO service page.