Your customer no longer asks Google, they ask ChatGPT, and the answer names only three companies. protan.ai measures why yours is not one of them, compares you with your competitors in the same scan, and lists what it takes to get into that answer.
| # | Company | Visibility | Position | Sentiment |
|---|---|---|---|---|
| 1 | Halstead Stainless | %63 ▲ 56 | 1.8 | |
| 2 | Ashcroft Metalworks | %57 ▼ 3 | 2.1 | |
| 3 | Northgate Stainless | %40 | 2.6 | |
| 4 | Redmont Steel Fabrication | %23 ▼ 9 | 3.1 |
* An example screen from the dashboard; figures and company names are illustrative.
Coming tenth in search results was bad, but you were still on the page. An AI answer names three companies, and there is no fourth. That is the difference.
Named in 2 of 30 prompts · Visibility 12/100
Named in 19 of 30 prompts · Visibility 71/100
The example above is illustrative. Real results depend on the sector, on what the company already publishes and on how quickly the work is carried out.
The score is a map, not a grade. Every component is tied to a specific fix, so the answer to "why is it low" is always a piece of work.
How often, and in which position, AI names you in the prompts real buyers use.
Is the business it describes your business, or has it quietly drifted into a competitor's story.
Can AI read your site at all. robots.txt, llms.txt, schema, server-side rendering.
The evidence it looks for before putting your name in an answer: third-party records and mentions.
The two purple components are calculated without asking the AI assistants anything, by reading your site and public records; the free scan shows exactly these two, 45% of the total weight. Both are independent of the AI engines and are calculated in full in every package. The two green ones require asking the engines in your package the prompts of your own sector, and that is where the packages differ.
Four tabs, four screens from the dashboard. All of them drawn from the measurement's own data.
Four components sit under the single number and each one is tied to a fix. The answer to "why is it low" is always a piece of work.
One word per engine: recommends, merely knows, or never mentions. Underneath sit your hit rate and average position.
Every answer behind the score is stored. The report shows examples where you appear and where you do not, next to the competitor that was recommended instead.
The leading manufacturers in the region are Halstead Stainless, Ashcroft Metalworks and Northgate Stainless. Halstead Stainless documents its ISO 3834 certified welding process for food-grade tanks on its own site.
Each item is bound to a score component and an expected movement. You mark what you did; the re-measurement in monitoring verifies it.
AI answers are random: the same model gives two different answers to the same question. So we do not fix the measurement, we fix the way it is measured.
The prompts written for your sector are locked on the first scan and the same ones are asked again in every monitoring re-measurement. If the prompts change, the before-and-after comparison is void; "prompts refreshed every time" is a marketing line, not a measurement.
Looking at a single answer is rolling dice. The score is the share of repeats that name you, and the report always shows it with its variance range: not 71, but 71 ± 4.
Signed-out session, memory off, fixed location. Otherwise what you measure is not the answer the AI gives, but the history of the person measuring it.
In the same scan we also measure 2–3 competitors who are doing nothing. If everyone rose, the model changed; if only you rose, the work worked. This is the only honest way to tell the two apart.
Every report states in its imprint which method version produced it, on which date and with which prompt set. If we change the weights, older reports keep carrying the rules they were produced under.
Three steps answering a single question: did the work change the answer. The three panels below are exactly what you get at the end of each step.
The buyer prompts written for your sector are asked three times to every AI in your package. Every answer is recorded one by one; whether your name appeared, and in which position, is marked separately. On a five-engine package that comes to 300 separate answers. In the same week the machine-readable side of your site goes through 14 automated checks.
Each item describes one job: what to do, why, and how many points it is expected to move which component. The order is set by the weight of the gap found in the scan, not by our opinion; that is why the top of the list is usually a one-line fix.
Re-measurement runs with a monitoring subscription: same prompts, same AI engines, same method version. Beside you we also measure 2–3 competitors who are doing nothing: if you all rose, the cause is a model update and the report says so; if only you rose, the cause is the work. If the score drops, we write that too.
The chart is illustrative. Real change depends on the sector, on what the site already publishes and on how quickly the work is carried out.
None of the following is a promise of results. Each one is our commitment on how the result will be measured, and each one can be verified item by item in the report.
There is no "improvement rate" in this table. Because that is for the re-measurement in monitoring to say, not for us. What we commit to is not the result itself, but that it is measured with the same method before and after and that whatever comes out is written down as it is. If the score drops, we write that too.
The Structure and Authority components are independent of the AI engines and are calculated in full in every package. Both read your site and public records. The number of engines only sets the Visibility and Accuracy side. Each engine gets its own report, because ChatGPT's reason for not mentioning you is not Gemini's reason.
Measured on one AI engine
one-off
Buy the Starter packageWhat's included:
Measured on three AI engines
one-off
Buy the Growth packageEverything in Starter, plus:
Measured on five AI engines
one-off
Buy the Pro packageEverything in Growth, plus:
One-off payment · no automatic renewal · taxes calculated at checkout
If you give us FTP access, we write and upload the robots.txt, llms.txt, llms-full.txt, sitemap.xml files to your root directory. Only files are uploaded to your root directory; your site content and code are not touched. If FTP access is not given, the report is still delivered in full and the setup right stays open for 90 days.
Regular tracking on ChatGPT
monthly · cancel any time
$566 billed yearly · 20% off
What's included:
Frequent tracking on three AI engines
monthly · cancel any time
$1,334 billed yearly · 20% off
Everything in Monitor, plus:
Full tracking on five AI engines
monthly · cancel any time
$2,870 billed yearly · 20% off
Everything in Monitor Pro, plus:
Cancel any time · taxes calculated at checkout
The engines change every week. Automatic rescans, an alert when your score drops, competitor tracking. It is switched on optionally after you buy your package and stopped with a single click whenever you like; on cancellation it runs to the end of the period you paid for. There are two tiers: three or five AI engines.
Item-by-item comparison, monitoring tiers and billing terms →
It measures the Structure and Authority components; both are calculated by reading your site and public records, without asking the AI assistants anything. The same checks are run for your competitors too, so you see the comparison for free. Visibility and Accuracy require asking the engines the prompts of your own sector; that measurement is done in a paid package, and the package you choose decides how many engines it covers.
No. The Structure and Authority components are independent of the AI engines; because they are calculated from your site and public records, they are measured the same way, in full, in all three packages. The number of engines only sets how many engines the Visibility and Accuracy components are calculated across. A single-engine package is not an incomplete measurement but a narrow one: it gives a complete result for ChatGPT and says nothing about Gemini.
Because the reasons are not the same. ChatGPT's reason for not mentioning you may sit on the index side, while Gemini's may be a missing Google Business record or Wikidata entity, and Claude's may be that your site is not rendered server-side. A score that squeezes every engine into one average cannot tell you which job to do.
It may well have changed, and to tell the two apart we measure 2–3 of your competitors in the same scan. If you all rose, the cause is a model update and the report says so. If only you rose, the cause is the work. Any score that does not draw this distinction is telling you nothing about cause and effect.
No. AI engines change their indexes and answer policies without notice; nobody has the control it would take to guarantee that outcome. What we do commit to is this: the measurement before and after is made with the same method and the same prompts, and whatever comes out is reported as it is. If the score drops, we write that too.
Because asking 600 prompts once carries less information than asking 10 prompts three times. Model answers are volatile; asking a question once and calling the result a score is measuring noise. A small but repeated and frozen set is the only setup that makes a before-and-after comparison possible.
A free score in 60 seconds. No card required.