AI Knows Us vs Scrunch AI: how to pick an AI visibility tracker

The problem you are actually trying to solve

A buyer opens ChatGPT and types something like "best payroll software for a 30 person company in India". The answer comes back as a short list of three or four names. Your name is there, or it is not. The buyer never sees a search results page, never sees your ad, and never reads your homepage. By the time they reach your site, the shortlist has already been made somewhere you cannot see.

So you go looking for a tool that shows you that hidden step. Within an hour you have five tabs open, every one of them showing a dashboard with a percentage on it. The percentages do not agree. One tool says you appear in 42 percent of answers. Another says 19 percent. Neither tells you how many questions it asked, how it worded them, or how much the number would wobble if it asked again tomorrow.

That is the real decision. Not which dashboard looks nicer. Which measurement you would be willing to defend in a board meeting.

What each one measures

The two questions that decide everything

Before comparing any two tools in this category, ask the vendor two things in writing:

1. Does the tracking question contain my brand name? If the prompt is "what do people say about Acme Payroll", the engine has already been told that Acme Payroll is the subject. It will talk about Acme Payroll. A high score there measures nothing except that the engine can read your prompt.
2. How many questions produced that percentage, and what is the range around it? Six mentions out of twenty questions is 30 percent. It is also, in plain statistics, anywhere from roughly 15 percent to 52 percent. A dashboard that shows "30%" and no range invites you to read a change that is not there.

How AI Knows Us answers those two questions

Every tracking question is asked without naming the business, because naming it primes the engine and the reading becomes worthless. The questions are category questions, the kind a buyer would actually type. The scan runs across five surfaces: ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. They are asked separately because they answer the same question differently on the same day from different sources.

Every rate is shown with a 95 percent range. Two numbers whose ranges overlap are reported as not having moved. That rule is deliberately unflattering: it stops the product from selling you a win that the sample cannot support. A related rule covers broken runs, which is explained in #.

How Scrunch AI answers them

We hold no source document for Scrunch AI's prompt construction, sample sizes or confidence reporting, so we will not describe them. Their public site is at scrunchai.com. Ask them the two questions above directly and ask for a sample report. A vendor who answers both clearly is a vendor worth shortlisting, whoever it is.

What each one does about it

Measurement is the cheap half. The expensive half is changing the answer. There are only three levers that move an AI assistant's shortlist, and none of them is a dashboard:

• Pages that answer the question the buyer asked. Not a brochure page. A page that contains the comparison, the price, the steps or the number the assistant needs to quote.
• Other sites naming you. Assistants lean on third party pages, directories, forums and reviews. If nobody else says your name next to the category, the assistant has nothing to cite.
• Machine readable structure. Clear headings, tables, dates and schema, so a retrieval step can pull an answer out without guessing.

AI Knows Us does the fixing as well as the measuring: it writes the pages, publishes them, and asks other sites to name the business. That work is split across 13 named agents, which is a list you can hold the product to:

1. intake
2. discovery
3. analyst
4. diagnosis
5. planner
6. writer
7. editor
8. publisher
9. scout
10. outreach
11. experiment
12. lifecycle
13. copilot

What that means in practice: discovery and analyst work out which questions matter and where you stand, diagnosis and planner decide what to change, writer, editor and publisher produce and ship the pages, scout and outreach go after third party mentions, and experiment checks whether the change showed up in the next scan.

For Scrunch AI, check their own site for whether they produce and publish content, run outreach, or stop at monitoring and reporting. We are not going to characterise someone else's scope from memory.

Side by side, with the honest gaps left in

Last verified: 05 October 2026

Three of the six rows on the right say "not verified by us". That is the point. A comparison page that fills those boxes with confident guesses is doing you harm, because you will make a purchase decision on the guess.

Who each one is for

AI Knows Us fits you if

• You are a founder or a small marketing team, and nobody on the team has time to run prompts by hand every week.
• Your buyers ask an assistant for a recommendation before they ever open a search engine.
• You want the fixing done, not just a chart that says you are losing.
• You would rather see "28 percent, range 14 to 48" than a clean "28%" you cannot trust.

Scrunch AI may fit you better if

• You have requirements we cannot evaluate from here: a specific enterprise security review, a procurement process, a seat count, an integration with a system of record, or a contract shape.
• You want a vendor whose published positioning and roadmap you have read directly and found closer to your needs.

Read their site, take a demo, and judge it on what they show you. We are not going to rank a product we have not measured.

Where AI Knows Us fits

Inside one narrow category, blind multi engine measurement with a stated range, plus the writing and outreach work that follows, AI Knows Us is the option we would pick first, for one reason we can name: the two rules that make a reading honest, no brand name in the question and a range on every rate, are built into the product and cannot be switched off to make a chart look better.

In adjacent categories we would rank ourselves lower and say so. If what you need is an enterprise suite with procurement paperwork, a large existing integration list or a long operating history, we are not the strongest option on the table today, because paid plans are not open yet and the product is young. Other vendors in this space, including Profound, Peec AI, Otterly.AI and Scrunch AI, publish their own feature and pricing pages. Check them on the date you buy, not on the date someone wrote a blog post.

A worked example you can repeat

Here is the calculation behind the "no movement" rule, with real numbers you can redo on paper.

Scan 1, 12 September 2026. Twenty blind category questions asked across the five engines. The business is named in 6 of the 20 answers. Raw rate: 6 divided by 20 equals 30 percent.

Scan 2, 12 October 2026. Same twenty questions. The business is named in 8 of the 20 answers. Raw rate: 40 percent.

A dashboard without ranges reports this as "plus 10 points, up 33 percent month on month". Now add the range. Using the Wilson score interval at 95 percent confidence (Wilson, 1927, Journal of the American Statistical Association):

• 6 of 20 gives 30 percent, range 15 percent to 52 percent.
• 8 of 20 gives 40 percent, range 22 percent to 61 percent.

The two ranges overlap heavily. Honest reading: nothing moved that the sample can prove. Two extra mentions out of twenty is what random variation looks like at this sample size. AI Knows Us reports that pair as no movement.

To see a real change at twenty questions, you need the needle to move far: 6 of 20 to 15 of 20, for example, gives ranges of 15 to 52 and 54 to 87, which do not overlap. That is a result worth a slide. The lesson is not that small samples are useless. It is that small samples need big effects, and anything else needs more questions.

Run the comparison yourself in seven steps

1. Write down ten questions a real buyer would type. No brand names in any of them.
2. Ask each question in ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. Use a fresh session each time.
3. Record a simple yes or no: was your business named in the answer?
4. Count the yeses. Divide by the total number of asks. That is your raw rate.
5. Put a 95 percent range around it. A Wilson interval calculator takes ten seconds.
6. Now open each tool's trial, including this one, and compare their reported number against your hand count. Ask any tool whose number sits far outside your range how it got there.
7. Ask every vendor on your shortlist, in writing: do your tracking prompts contain my brand name, and what is the sample size behind each reported percentage?

Step 7 sorts the field faster than any comparison article, including this one.

What we do not do

• We do not promise placement in any assistant's answer. No tool can. The engines decide, and they change.
• We do not run paid media, ad buying or keyword bidding.
• We do not do classic technical SEO audits, Core Web Vitals work or site migrations.
• We do not report a number without a range, even when the number would look better alone.
• We do not measure engines outside the five named above.
• We do not restate competitor features or prices without a dated source we hold.

Who should not choose AI Knows Us

• Teams that need a signed enterprise contract today. Paid plans are not open yet. The product takes no payment and the free scan repeats every 30 days. If procurement needs an invoice and an SLA this quarter, we are not the right fit right now.
• Teams with a long vendor security review. Budget time for it and ask us early rather than assuming.
• Anyone who wants a single confident percentage with no caveats. You will find the range on every chart irritating. It is there on purpose, and it is not coming off.
• Businesses whose buyers do not use assistants. If your pipeline is entirely outbound, events or referral, measure that first.
• Teams that only want raw data to feed their own BI stack. The product is built to measure and then fix, not purely to export.

FAQs