The secret hacks are a pipe dream — because there's no technical trick that fakes the actual currency: being mentioned and cited by other people all over the web. That's earned slowly, or bought outright through the ad products the big platforms just switched on. An unknown with a clever llms.txt file and perfect schema is optimizing a door the machine barely looks at.
You built a good site. You've heard that AI is the new search, and somewhere in your feed a confident person is selling a checklist of "GEO secrets" to make ChatGPT name-drop your product. You want to know if it works. Short version above; the honest, sourced long version follows.
The whole "AI SEO secrets" genre rests on one unspoken promise: that there's a switch on your site you can flip to make the models favor you. Add the magic file, sprinkle the structured data, and the red carpet rolls out.
It doesn't. The clearest evidence is a controlled test from Ahrefs: they took 1,885 pages, added structured data (schema markup), and measured AI citations before and after. The lift across ChatGPT, Perplexity, and Google's AI answers was, essentially, nothing. The single most-sold "GEO hack" moved the needle zero.
That's the pattern under all of it. The tricks people sell are the parts of the system you control — your markup, your keywords, a text file in your root. And those are exactly the parts the models weight the least. The machine isn't reading your site looking for permission to recommend you. It's reading the rest of the internet to decide whether you're worth mentioning at all.
The hacks sell because they're the only part you control. The models ignore them for exactly the same reason.
There are only two ways your product's name ends up inside an AI answer. One is paid. One is earned. Neither is a hack, and the paid one is exactly the SEO-and-search-ads story you already suspected.
Your instinct that Big AI would monetize this the way Google monetized search was correct, and it happened faster than most people noticed:
This is the red carpet, and it works precisely like paid search: whoever funds it, wins the slot. If you have the budget of a venture-backed brand, this door is open. If you're a solo operator, it's a toll road you can't afford to live on.
The organic door is real, and this is the part that surprises people: it is not dominated by whoever is biggest. It's dominated by whoever the wider web is talking about. Across a study of 75,000 brands, the strongest signal for showing up in Google's AI answers was brand web mentions — correlation 0.664, versus a feeble 0.218 for backlinks. Being talked about beat being linked to by roughly three to one. The brands in the top tier of mentions earned up to 10× more AI mentions than the next group down.
And traditional ranking doesn't carry over the way you'd hope. In one documented case, a law firm sitting at #1 on Google for its main money keyword got zero mentions in ChatGPT. Winning the old game buys you nothing automatically in the new one.
Sorted honestly. The left column is where your effort pays. The right column is where it's being harvested by someone selling a course.
| Actually moves the needle | Sold as a secret — mostly doesn't |
|---|---|
| Earned mentions. Being named in articles, reviews, forums, and industry chatter. The single strongest signal there is. | Schema / structured data as an AI hack. Tested on 1,885 pages: zero citation lift. (Still worth doing for normal search — just not the magic bullet it's sold as.) |
| Reddit and community presence. Reddit alone accounts for a quarter to nearly half of what some engines cite. Real participation, not spam. | An llms.txt file. A nice-to-have at best; no engine has committed to it as a ranking input. It won't rescue a site nobody mentions. |
| Original data and research. Genuinely new numbers get cited about 82% of the time; generic blog rehash, about 25%. | Keyword density and "AI-optimized" copy. Writing for the model instead of being worth citing. The model can tell the difference now. |
| Freshness. Recent, dated content wins. Even a visible "2026" in the title tracks with roughly 30% more citations. | Buying backlinks. Links correlate far weaker than mentions (0.218 vs 0.664). The old link game barely transfers. |
| Being fetchable. The top-scoring technical factor of all. If a crawler can't read the page, nothing else matters. | Accidentally blocking the crawlers. Many sites quietly block the AI bots in robots.txt and never know — then wonder why they're invisible. |
Here's where the "combine web-page and server coding" question actually earns an answer — because the technical work does matter, just not the way the hack-sellers frame it. The code layer isn't how you win. It's how you avoid disqualifying yourself. Get it wrong and the best content on earth never gets read by the machine.
robots.txt and any firewall/CDN rules for GPTBot, OAI-SearchBot, PerplexityBot, Google-Extended and friends. Blocking them — often a default someone flipped for "security" — is the most common act of self-sabotage on the web right now.Do the schema markup too — but do it for the humans and the normal search engines who still send you most of your traffic, not because you believe it's the secret handshake. It isn't.
Make it trivially easy for a machine to read your words, then get out of the way. That's the entire job. Everything past "readable, fast, unblocked, current" is diminishing returns dressed up as a strategy.
You can't hack it, and you can't outspend a funded brand on the paid lane. That sounds like a dead end. It isn't — it just means the only move left is the one nobody's selling as a shortcut, because it isn't one.
Stop trying to be optimized. Start being worth citing. The machine mentions you when the web mentions you, so the entire game collapses to a single question: why would anyone talk about you? Manufacture real reasons.
None of that is fast, and none of it fits in a checklist you'd pay $49 for. That's the point. It's the same pattern that has always rewarded the small and unfunded: make one genuinely useful thing, let the right people notice it, and let that noticing compound. The machine is just a new set of eyes on a very old game.
So, is getting Big AI to promote you a pipe dream? The secret-hack version absolutely is — and the sooner you stop paying for it, the better. The funded already have their paid express lane, and no amount of clever markup gets you into it for free.
But the earned lane is not a pipe dream — it's just honest work wearing an unfamiliar costume. AI didn't invent a new way to win. It rewired the plumbing so that being genuinely talked about matters more than ever, and being technically clever matters less than the courses claim. Make something worth mentioning, make it trivially readable by a machine, and get mentioned by real people in real places. That's the whole strategy. Everything else is somebody selling you a map to a door that doesn't open.
The brutal truth was never that you can't get found. It's that there's no way to skip the part where you earn it.
No secret-hack checklists. Just what I find when I actually test the claims — from someone running a business on this stuff, not selling a course about it.
Unsubscribe anytime. Your email stays private.
— Scott Covert. If you've been sold one of these "GEO secret" checklists — or you've found something that genuinely moved your AI visibility — I want to hear it. Reach me here.