How to Spot a Fake AI-Visibility Expert or Tool
Opening
AI visibility is the new gold rush. Every week another "AI-visibility expert" shows up promising to get your store recommended by ChatGPT. Some are real. A lot are the old search-gaming crowd wearing a new hat. And some are not people at all, but software that promises to do the whole thing for you, no expertise required, which is the more tempting trap because it looks like empowerment. Person or platform, a few of them are selling something that cannot be delivered by anyone.
I know the difference right now for a specific reason. This month I built the honest version: a measurement system that tracks whether the work actually moves your citations, and it does that by comparing pages I worked on against pages I left alone, so I can tell what my work did versus what the AI engines did on their own. Building the real thing made the fake thing obvious. Once you know what earned citations look like and how slow and structural they are, the shortcuts and the guarantees stick out like a sore thumb.
You do not have to become technical to protect yourself here. You just need to know the tells. Here they are.
Key takeaway
Real AI-visibility work is structural, measurable, and never guaranteed. It makes your store genuinely easier for AI to read, trust, and cite, and it proves what it moved. Fake work tries to trick the machine, cannot be measured, and promises outcomes no one controls. If a provider cannot show you both their method and their measurement, walk away.
Why this matters right now
AI visibility is new, it is unregulated, and it is suddenly worth money. That is the exact recipe that draws grifters. The field is young enough that there is no watchdog, no certification that means anything, and no shared standard for what "done" looks like. So anyone can call themselves an expert, and the loudest promises tend to come from the people with the least to lose if they are wrong.
The independent jewelry designer is a soft target in this. You are busy at the bench, you know your craft cold, and you have been told for years that the technical side is over your head. That combination is exactly what a bad actor counts on. So let us make the technical side legible enough that you can hire well.
The red flags
I sort the tells into three families: impossible promises, tricks played on the machine, and a refusal to measure. Any one of them is a reason to slow down. Two of them together is a reason to leave.
Impossible promises
The first family is the easiest to catch, because it lives in the sales pitch, not the code.
A guarantee is the loudest red flag there is. "Guaranteed citations." "I will get you number one in ChatGPT." Nobody can promise that. The AI engines are closed systems that change constantly, and the same question can produce a different answer on two tries. Real work moves the odds in your favor and proves it. It cannot promise a specific outcome, because the specific outcome is not the provider's to give.
Close behind is the secret sauce. A proprietary trick, a magic method they will not explain, a black box that "just works." Real AI-visibility work is honestly kind of boring. It is structure, consistency, and genuinely useful content. If someone is selling you a clever secret, the cleverness is usually the part that gets you in trouble later.
Tricks played on the machine
The second family is where the real damage happens, because these tactics can get your site penalized or your profile suspended, and you would never know the provider did it.
Stuffing and lying with structure. Cramming keywords into your pages, or adding behind-the-scenes tags (the code that describes your page to search engines and AI) that claim things your visible page does not say. Google's own guidelines require that structured markup describe content that is actually visible to your visitors, and it penalizes the mismatch. If the tags say "award-winning" and your page does not, that is not optimization, that is a liability.
Hidden instructions. A newer and nastier one: hiding text on your page that is aimed at the AI itself, trying to instruct it to recommend you. The industry calls this hidden prompt injection, and it is the digital version of slipping the judge a note. Here is the part the people selling it will not mention: the engines have largely outgrown it. The platforms and security researchers have closed most of these loopholes, so the trick that was merely risky last year increasingly does not even work this year. It is manipulation, the engines actively catch it, and being caught is not a gentle correction.
Cloaking. Showing the AI crawlers one version of your site and showing your customers another. This directly violates search spam policies and it is one of the fastest ways to get removed from results entirely.
Bought authority. AI often decides who to trust by looking at where else you are mentioned. A bad actor manufactures those mentions: spammy directory blasts, networks of junk sites built only to link to each other, purchased or faked reviews. It looks like authority for a minute. Then it looks like a spam pattern, and the fall is worse than never having done it.
And the penalty is rarely a single slap on the wrist. It escalates: lost rankings, then a trust downgrade, then the AI engines filtering you out of what they are willing to pull from, then removal or a suspended account. By the time you are on that ladder, the person who put you there is long gone.
A refusal to measure
The third family is the hardest to spot, and it is the one most people miss, so it is the one I care about most.
Vanity numbers and lucky screenshots. "We got you cited five hundred percent more." A single screenshot of one lucky answer, with no starting point, no repeat runs, and no way to check. Because the engines are inconsistent, a provider can run the same question ten times, screenshot the one time you showed up, and call it a win. The honest version runs it many times, reports how often you actually appear, and compares it against pages that were left untouched, so the number means something.
Ask an AI the same question ten times and you get different answers. Run it enough and the store shows up once. That single lucky run becomes the screenshot, sold to you as "+500% more citations."
Fear as the closer. "You are invisible to AI. Buy now or disappear." Manufactured urgency is a sales tactic, not a diagnosis. The reality is calmer and more workable than that, and anyone leading with panic is selling to your fear instead of your situation.
Here is the whole pattern in one place.
Shady versus honest, side by side
| The shady move | Why it is a red flag | What honest work looks like |
|---|---|---|
| Guarantees a citation or a number-one spot | The engines are closed and change constantly; no one controls the outcome | Moves the odds in your favor and proves the movement |
| Sells a secret method or magic trick | The cleverness is usually the part that causes harm | Explains the method plainly; the real work is structural and unglamorous |
| Stuffs keywords or adds tags that lie about the page | Violates Google's structured-data policy; penalizes the mismatch | Makes the visible page genuinely clear, then marks up what is really there |
| Hides instructions on the page aimed at the AI | Manipulation the engines actively catch, and increasingly ignore | Earns the recommendation with real, useful content |
| Shows crawlers a different site than customers (cloaking) | Direct spam-policy violation; fastest route to removal | One honest site for people and machines alike |
| Manufactures mentions with spam directories or bought reviews | Reads as a spam pattern; the fall is worse than never trying | Earns mentions by being genuinely worth referencing |
| Reports one lucky screenshot as proof | No baseline, no repeat runs, no control; the number is theater | Measures frequency across many runs against untouched pages |
| Closes with fear and urgency | Selling to your panic, not your situation | Gives you the honest picture and a workable next step |
The software trap: when the magic is a token meter
Not every bad actor is a person. A lot of them are software, and software is the more tempting trap, because it looks like empowerment. Run our tool, no expert needed, watch your visibility climb. It feels like doing it yourself, which is exactly why it works on people who have been told the technical side is over their heads.
I almost learned this one the expensive way. I got swept up in an AI-visibility platform that dazzled everyone at its launch. Beautiful demos, an agent that promised to do the work for me, a feature for building authority with backlinks. I bought a stack of credits and dove in. Here is what I found once I was inside.
The "link building" bought placements from a set of partner sites that were blurred out, so I could not see which sites my brand was about to be linked from. That is not authority. That is a spam directory with a friendly interface, pointing junk at your name.
The pricing was its own trap. A modest monthly fee up front, then hundreds of dollars in credits you have to keep buying to actually use the thing. The features seemed built to route me toward spending more. And when I tried to leave, I learned the work does not come with you: disconnect, and it is gone. You pay forever, or you lose everything.
When a bug drained my credits and a promised course never arrived, support ignored my ticket. What finally moved them was going public: a comment on a forum, and a review honest enough to dent their rating. Only then did anyone respond, and even then I recovered just part of what I was owed. I happen to know where to push to get heard. Most people do not, and in that platform's own reviews, plenty of them were not as lucky as me. I got out fast, and I am grateful, because the alternative was paying more and watching a pile of spammy links drag down a brand I had spent years building.
So aim these red flags at software just as hard as at any consultant:
- The real price is hidden behind credits or tokens. A cheap base tier, then hundreds more to make it actually work. The base fee was bait.
- The "link building" will not show you the sites. If you cannot see who is linking to you, assume it is junk.
- You are locked in. Cancel and lose the work. Real improvements to your own site stay with you. If leaving erases everything, you never owned it.
- The features are built to upsell you. When the tool's job is to route you toward spending more, its job is not your visibility.
- There is no real recourse. If getting made whole requires a public fight, that is your answer before you ever pay.
Magic software that promises AI visibility with no work and no expertise is the same promise as the guarantee. It cannot be true, and the version that pretends it can is usually the version that hurts you.
Check it yourself, for free
You do not have to take anyone's word, and you do not have to be technical. Here are free tools that let you check the things you cannot see, whether the culprit is a consultant or a piece of software.
Start with the two that need no login at all:
- See what your page actually claims. Paste your page into Google's Rich Results Test or validator.schema.org. It shows the behind-the-scenes tags on your page, so you can catch tags that claim things your visible page does not say.
- Reveal hidden text. Open your page and press Command-A or Control-A to select everything. White-on-white text hidden for the AI lights up when it is highlighted.
Then, if you already have Google Search Console connected, it is your best friend here. Setting it up is the one genuinely fiddly step in all of this, so if you do not have it yet, treat this tier as optional, not homework:
- Did someone hurt my site? The Manual Actions and Security Issues reports tell you plainly if Google has penalized your site. This is the single most important check after any provider or tool has touched it.
- Is a pile of junk pointing at me? The Links report shows who links to your site. A sudden wave of irrelevant or foreign junk domains is the spammy-backlink red flag, exactly the risk of those blurred link directories.
- Does Google see a different page than my customers? Use URL Inspection, then View Crawled Page, and compare it to your live page. A big mismatch is a cloaking flag.
None of this requires you to read code. It requires ten minutes and a willingness to look.
How to vet an AI-visibility provider
You do not need to audit their code. You need to ask five questions and listen to how they answer.
- "How will you measure whether this actually worked?" If the answer is a vague promise instead of a method, that is your answer. Look for words like baseline, repeat runs, and comparison.
- "What can you NOT promise me?" An honest expert will happily tell you what is outside their control. Someone who claims control over everything controls nothing.
- "Show me the change on the page, not just in the code." Real work changes what a human can read on your site. If everything is invisible and behind the scenes, ask why.
- "Are the mentions you build earned or bought?" Earned takes longer and lasts. Bought is fast and fragile. You want the slow kind.
- "What happens if the AI changes its mind next month?" The honest answer is that structural work holds up because it is built on being genuinely good, not on a loophole. If their whole plan is one trick, one update erases it.
FAQ
Can anyone actually guarantee I will get cited by ChatGPT?
No. The AI engines are closed systems that change often, and the same question can return different answers on different tries. Good work improves your odds and proves the improvement. It cannot promise a specific placement, and anyone who does is either uninformed or dishonest.
Isn't schema just adding some code? How can that be shady?
The code (the behind-the-scenes tags that describe your page to search engines and AI) is fine and useful when it describes what is actually on your page. It becomes a liability when it claims things your visible page does not say. Google's guidelines require the two to match, and it penalizes the mismatch. The shady part is not the tags. It is lying with them.
How do I know if my current provider is doing any of this?
Ask them the five vetting questions above, especially the measurement one. Then look at your own site: is there hidden text, are there tags claiming things your pages do not, did a pile of low-quality mentions appear all at once. If you are not sure what you are looking at, that is a fair thing to have a second set of eyes check.
Could shady tactics actually get my site penalized?
Yes. Cloaking, hidden manipulation, and faked authority all violate published search policies, and the penalties range from lost rankings to removal. The cruel part is that the person who did it to you is long gone, and you are the one left cleaning it up. That is why the boring, honest version is also the safe one.
Sources
- Google Search Central, Structured data guidelines: structured markup must describe content visible to users.
- Google Search Central, Spam policies: cloaking and scaled content abuse are violations.
- Search Engine Land, Black hat GEO is real, here is why you should pay attention: industry reporting on AI-search manipulation tactics and their penalties.
- Digiday, GEO hype busted: experts call it more SEO than a new discipline: why the real work is structural fundamentals, not a secret new field.
- Similarweb, Black Hat GEO: tactics and risks: a catalog of manipulation tactics and their escalating consequences.
- Andrea Li, building Red Pin Geek's AI-citation measurement system, July 2026 (first-party).
Continue Learning
- Why AI search skips independent jewelers: the Foundation Gap: the real work is foundational, which is exactly why the shortcuts fail.
- How ChatGPT decides which jewelry store to recommend: what the engines actually reward, which is the opposite of gaming them.
- Score your own store on the free AI Visibility Score: before you hire anyone, see your real starting point, so you can tell honest progress from theater.
- Subscribe to the Red Pin Geek Substack (free): weekly inside-view content for jewelry designers building AI visibility on their own. The Premium tier unlocks the scorecards and worked walkthroughs behind these checks.
- Work with me on your specific store (Snapshot / Full Audit / Full Audit + 1:1 call): the private read. I run the same honest checks from this piece on your store and tell you what I find, including whether a past provider left anything risky behind.
Author bio
Andrea Li is the founder of Red Pin Geek and has spent 18 years in the jewelry industry. She helps independent jewelry designers get found and recommended by AI search, honestly and measurably. More at redpingeek.com and andreali.com.
Want Red Pin Geek in Your AI Answers?
Google now lets you choose preferred sources that get labeled in AI Overviews and AI Mode. If this content helps you navigate AI visibility for your jewelry business, add us and you'll spot our links faster next time Google answers your questions.
Add Red Pin Geek as a Preferred SourceTakes one click. Works in Google Search, AI Overviews, and AI Mode.
