Key Takeaways
- Make every claim recomputable before you pay. A score you cannot recompute is a number somebody chose. Ask how it was calculated, and treat an unwillingness to explain it as the answer.
- Refuse any promise the vendor does not control. Nobody controls what an assistant says. A guaranteed citation by a fixed date is either a misunderstanding of the product or a sentence written to close you.
- Demand evidence behind every competitor name. A list of rivals is an assertion you can only agree with. The quoted sentence that produced each name is the thing you can actually check.
- Get the price in writing before the call. A price that only exists in conversation is a price set by how you sound on the phone. That is a fact about their sales process, not about your business.
- Ask what the day-90 number gets compared against. A before and after only means something when the same frozen questions were asked both times. If nothing was stored, there is nothing to compare.
The AI visibility category is about two years old, which means the honest operators and the opportunists are still using the same vocabulary. From the outside, the pitches look identical.
They are not identical underneath, and the difference is checkable. Every flag below can be settled with one question, before any money moves.
We sell in this category, so read this the way you would read anything a vendor writes about its competitors. The last section is the part where we apply the list to ourselves.
What are the red flags in an AI visibility vendor?
Nine, and they fall into three groups: how the vendor measures, what the vendor promises, and how the vendor contracts. A single flag is a question. A flag from all three groups is a pattern.
None of them requires you to understand how a language model works. Each one is answerable by a buyer with ten minutes and a willingness to ask a direct question and wait through the pause.
The nine flags, and the test for each
Print this, or keep the tab open during the call. The right-hand line of each tile is the whole test.
One flag is a question worth asking. One from each group is a pattern, and the pattern is what should stop you.
The measurement group is the one people skip, because a number in a report looks like a fact. It is worth being blunt about why a single run is not one. We ran one identical set of 25 questions twice, an hour apart, on August 1st 2026, with nothing changed in between, and the two lists of rival businesses had no overlap at all. Carry the limits of that with it: one question set, one niche, one day, and our own test rather than a published study. It is enough to show that a single answer is one roll of a die, printed as a finding.
The two promises nobody can make
Two of the nine are worth separating out, because they are not sloppiness. They are claims about things the vendor does not control.
The first is a guaranteed citation by a fixed date. Nobody owns the assistant. Perplexity's help center describes a product that searches and cites sources for every answer, and Google's guidance on its AI features describes a system running several searches behind one question. Both are systems that change without telling anybody. A vendor promising a specific mention on a specific date is promising something they cannot deliver, and the only question left is whether they know it.
The second is output priced as a result. Pages per week, posts per month, listings submitted. Those are measures of the vendor's activity, and a business can buy a great deal of activity without ever being named in an answer. The Princeton-led study GEO: Generative Engine Optimization found that what changed a source's visibility in generative answers was the content of the page, specifically things like citations, quotations and statistics, rather than the quantity of pages. Volume is the wrong unit.
A guarantee can be honest. It just has to be about the vendor's own money and the vendor's own work rather than the assistant's behavior. Ours is one net-new citation in 90 days or 90 more days free, and nothing at day 180 means a full refund. That is a promise about what we do when it does not work, which is the only kind anybody can keep.
What to ask, and what a good answer sounds like
Here is the same nine, written out with the answer you are listening for. A good vendor answers these quickly, because they have answered them before.
| The flag | What to ask | A good answer sounds like |
|---|---|---|
| A score out of 100 with no stated method | How is that number computed, and can I recompute it? | It is a count: named in N of the questions we asked, and here are the questions |
| One run, printed as fact | How many passes did you run, and what did the second return? | Two independent passes, and we only report what both agreed on |
| One assistant, sold as your visibility | Which tools were used, and can I see each one separately? | Five, reported per tool, because they disagree with each other |
| A guaranteed citation by a fixed date | Who controls the assistant that would have to say it? | Nobody does, so our guarantee covers our work and our money instead |
| Output measured in pages per week | What result is that number supposed to produce? | We do not sell volume, we sell one page that answers your worst question |
| Competitor names with no evidence attached | Can I see the sentence each name was judged from? | Every name carries the quoted line the assistant said it in |
| A price that only exists on a call | Can you send me the number before we speak? | It is published, and it is the same for everyone |
| A long commitment before anything is delivered | What lands in week one, and what if it does not? | Here is the week-one deliverable and here is the refund position |
| No re-run of the same questions later | What will the day-90 comparison be measured against? | The same frozen question set, stored at the start for exactly that |
Notice that none of the good answers are technical. Every one of them is a vendor describing their own process in plain words, which is the actual test hiding inside all nine.
Where we would trip our own list
Run the list against us and it catches something, so here it is before somebody else says it.
Our free check is one tool and one pass. It asks ChatGPT 25 buyer questions and reports what came back. By flags two and three above, that is exactly the thing we just told you not to accept as an audit, and it is not one. It is a baseline: enough to show you whether you are in the answers at all, and free precisely because it cannot carry more weight than that. If you treat a free check from anybody, including us, as a measurement you can act on for months, you are making the mistake this post is about.
The honest version of the rest: our prices are published at $0, $500, and $399 a month to keep watching, and the done-with-you work is quoted on request because the scope genuinely varies. We do not guarantee a citation by a date, and we would not take the money if we did. Plenty of businesses read a free readout, find one obvious gap, fix it themselves and never buy anything, which is a fine outcome and happens regularly.
The positive version of this list, the eight things an audit should contain, is in what a real AI visibility audit should include. What the category actually charges is in what AI visibility costs, and the choice between watching the number and changing it is in monitoring tools versus done-for-you work.
If you want a baseline before you talk to anyone at all, that is the sensible order. The free check asks ChatGPT 25 real buyer questions from your niche at $0 and emails a plain-English readout: your score, who got named in your place, and the one fix to start with. It runs as soon as you ask for it, and the report lands within the hour between 7am and 5pm Central, or by 10am the next morning outside those hours.
Questions we hear the most
What are the red flags in an AI visibility vendor?
Nine, in three groups: a score with no method, a single run, and a single tool; a guaranteed citation, output measured in pages, and competitor names with no evidence; a price only given on a call, a long lock-in, and no stored baseline.
Is a guaranteed citation ever legitimate?
No. No vendor controls what an assistant says about you, so nobody can guarantee a specific mention by a specific date. A guarantee can only honestly cover what the vendor does if the result does not arrive.
Why is a single-run audit a problem?
Because assistants are not deterministic. We ran one identical 25-question set twice, an hour apart, on August 1st 2026, and the two lists of rival names had no overlap. One run cannot tell a real competitor from a fluent guess.
What is wrong with being quoted a price on a call?
Nothing, if the number is the same for everyone. The flag is a price that exists only on a call, because a number that cannot be written down before the conversation is a number set during it.
Should I avoid monitoring tools entirely?
No. A monitoring tool that tells you your score honestly, cheaply and repeatedly is a reasonable purchase. The flag is a monitoring tool sold as though it also does the fixing, which is a different job.
What should a guarantee actually cover?
The vendor own work and the vendor own money. Ours is stated against a named result: one net-new citation in 90 days, or 90 more days free, and nothing at day 180 means a full refund.
How do I check a vendor claim without buying anything?
Run the free check first and read the baseline yourself. It asks ChatGPT 25 real buyer questions from your niche and emails a plain-English report: your score, who got named in your place, and the one fix to start with. It costs $0.
What does the $500 AnswerHalo audit do differently?
It runs 125 checks across five AI tools as two independent passes, names a rival only where two tools and both passes agree, shows the sentence behind each name, and ends with one specified page. It arrives by this time tomorrow.
Your buyers are already asking. Find out what AI tells them.
The free check asks ChatGPT 25 real buyer questions about your niche and sends you the report within the hour, at $0. If it shows you are already getting named everywhere, we will say so in plain words.
Get my free check →