Platform · Measure
We hold ourselves to a Higher Standard.
Five of them, in fact.
Every paid AEOSearch report is held to these 5 standards, each of which differentiates us from our competitors. Without these 5 data points, you cannot be ranked accurately with a visibility score. Any competitor who does not offer these isn't living up to our standards on transparency and quality.
- 01
Sample Size
We divulge every answer we count.
We clearly identify how many passes we are offering in your report, and we give you the choice of 1 or 3, though we will also offer custom reports for enterprise and agency clients. We recommend 3 passes because it will smooth out the noise of a single pass of answers, which happens with every LLM provider: there will not be a static, constant answer to any question, so asking the same question 3 times will provide more data which we can average for you.
- 02
Model Named
We allow you to choose which models you want to use for your report.
This transparency matters to us because most competitors shroud which engines they are using to provide you with the answers you pay for, meaning that they pay for the least used, cheapest models and claim that you are receiving data relevant to your business. We allow you to choose, because your business may dictate which models are relevant to your customer base. For instance, if you are an architect or marketing agency your clients are more likely to use newer frontier models, however if you are a mattress company or roofing contractor your clients may use older models because they don't expect the model to do extensive research to provide them with the answers they seek. These are generalizations, however we want to allow you to choose the models which you think are best for your search, which also dictates the price of your report.
- 03
Grounding Flagged
We tell you when an answer is grounded.
Grounded answers use web search to return their results, and they are more expensive because they use more tokens and energy with each question asked. We turn on grounding for our question sets because we want to provide you with the best report possible, and because we think it is unfair not to explain this to our customers. Ungrounded answers generally present dated and inferior information, while grounded answers should provide you with contemporaneous and relevant data that would otherwise be missed.
- 04
Range Shown
We print the range of variation for every number we report.
A mention rate is a count over the questions we asked, so it is never exact because it can't be: For instance, even if ChatGPT named you in 8 of 20 questions in one pass, your mention rate is not always 40% because there will always be variation each time we ask the same 20 questions, so we will give you the expected range, which we calculate from every answer we collected for your report. If we run a 50 or 100 question report, the additional data will give all parties more confidence in your mention rate, as does asking the same question 3 times instead of just a single pass. These are all factors you should consider when you decide which report you want to order because the additional data leads to more confidence in your mention rate and tightens the range we can confidently present to you.
- 05
Every Source is Cited
Every number we present to you is catalogued and identified in our system.
We present you with the full text of the answers to the questions you approved, with their citations and the source provider. We expect to deliver you comprehensive and complete reports, so we will not charge you for answers which are not returned due to outages.
When a page is changed
Measuring is free to read and changing a site costs three more standards. A change missing any one of them is a defect, not a feature.
Previewed
The WordPress plugin and the Shopify app show the exact change before it lands; the platform itself never writes to a site.
Receipted
Every applied change is recorded with a rollback token, bound to the site that opted in, and the report says “applied, rollback available” beside it rather than “nothing has been changed.”
Reversible
A change that cannot be undone is refused before it is recorded, and revoking consent can never block a rollback.
What we refuse to print
- Never a promised mention or a projected ranking. AI answers change between runs, and every figure is a dated snapshot.
- Never an outage sold as invisibility. A failed call is stored as a failure, excluded from every rate, disclosed on the report, and not charged for.
- Never a crawl count we did not fetch. The pages a report says it read are the pages it read.
- Never an inferred AI-traffic number. Google AI Overviews and AI Mode visits arrive as Google organic and cannot be separated from it, so they are reported as exactly that.
Against the field
On 12 September 2026 we read every AI-visibility listing in the Shopify App Store we could find, 37 in all, for the same five standards. The column on the right is what those listings state, counted, not what we think of them.
| Standard | AEOSearch | 37 App Store listings |
|---|---|---|
| Model version stated | From each reply, wherever the provider names it | 0 of 37 listings |
| Grounding stated | Per engine, per answer | 0 of 37 listings |
| Sample count stated | On every report | 2 of 37 listings |
| Interval published | Beside the mention rate, overall and per engine | None found |
| Raw answer kept | Every answer, until you ask us to delete it | Not checked by the sweep |
Listings were read from their public store pages. A listing that keeps one of these standards inside the app rather than on its page would not be counted here, which is a limit of the sweep, and the reason the column says “stated”.
The same five standards on the daily read
The Consumer View is one read a day of the answer a signed-out consumer is served, and it is the least controlled measurement we sell. The standards still hold on it, and they are why it costs $27 a month rather than a report's price:
- Sample size: one. Every count on the page says so, and one read is never described as three.
- Model: as the surface reported it, printed beside each answer, never one we chose. Where the surface names no model, the field says that.
- Grounding: as the surface reported it. Where it reports nothing, the flag reads unknown, not off.
- Range: none. One read carries no interval, so the page prints a line of days and no percentage with a decimal point.
- Every source: the answer is kept, and shown under the count it produced.
The API against the consumer apps
When your report asks ChatGPT or Claude, it asks through their APIs, the access each company sells to software like ours, and not through the apps your customers open. The two don't always name the same brands, so once a month we ask the same questions both ways by hand and publish the gap. For the app side we use the ChatGPT, Claude and Perplexity apps on our own accounts. On the October 2026 payroll software panel (10 questions, 3 runs, 20 brands), how often the API named a brand differed from how often the consumer apps named it by 6.5 points on average, and by 10 points or less for 17 of the 20 brands.
The apps don't agree with each other either, so we don't expect the API to match them exactly: we hold the ChatGPT API and the Claude API each to how far the apps sit from each other. On that panel:
- The apps against each other: 8.6 to 11 points on average, depending on which two you compare, and 9.6 averaged across all three pairs. That 9.6 is the bar.
- The ChatGPT API against the ChatGPT app: 8.4 points on average, under the bar.
- The Claude API against the Claude app: 10.2 points on average, over the bar.
An engine over the bar puts its category on the following month's panel. We don't hold the API's figure above to the bar, because it pools both engines, and pooling flatters it twice: one engine's overcount can cancel the other's undercount, and more answers shrink the part of any gap that is only chance.
Being named is one scoreboard; where and how is another. Over the 19 brands both surfaces named, the API placed a brand 5.2th on average against 5th in the apps, and recommended it in 19% of answers against 18%. The same questions through the vendor captures the Consumer View buys (ChatGPT and Gemini) sat 7.9 points from the API and 5.2 from the hand-run apps.
Where the API sat more than 10 points from the apps on the October 2026 panel, in points of mention rate (negative means the API named the brand less often):
| Brand | API | Apps | Gap | Rank API / apps | Recommended API / apps |
|---|---|---|---|---|---|
| OnPay | 85% | 50% | +35 | 3 / 3.3 | 72% / 38% |
| Patriot | 58% | 41% | +17 | 4.2 / 3.7 | 42% / 27% |
| Paychex | 58% | 44% | +14 | 4.5 / 4.5 | 15% / 22% |
This is one panel of 10 questions in one category, 3 runs, so a brand's gap is one month's reading, not a law.
The one-minute version
An AEOSearch report tells you how many passes it counted, which model gave each answer, whether that answer was grounded, how wide the range around every number is, and where every number came from, with the full answer and its citations kept. When a page is changed, the change was previewed, receipted and can be undone.
The long version is the methodology. The free scorecard carries the same cover and the same 95% range on its mention rate as a paid report, plus the stored answer for up to three questions you lost, so the cheapest way to check us is to run one.
Start with the free scorecard.
Let's run the free scorecard and read the five standards on your own numbers.
$0 to start · no card required · every number traces to a stored answer