AI Knows Us vs Otterly.AI: an honest comparison for AI visibility tracking

The problem this page is about

A buyer opens ChatGPT and types "best payroll software for a 40 person company in India". Four names come back. If yours is not one of them, you never learn that the conversation happened. There is no referrer, no keyword report, no impression count. The sale was lost in a room you cannot see into.

That is why a whole category of AI visibility tools appeared in the last two years. They all promise the same headline: we will tell you whether the assistants mention you. The honest question is not which tool has the prettiest dashboard. It is this: can I trust the number, and does anything change after I read it?

Two tools get compared a lot for this job. One is Otterly.AI. The other is AI Knows Us, which publishes this page. We are not neutral. So this page is built so you can check it: the method is written out, the one test that separates good measurement from bad measurement is reproducible in ten minutes with a free ChatGPT account, and the claims we cannot confirm about the other product are simply not made.

What each one measures

The question that decides everything: does the prompt name you?

Every AI visibility tool sends prompts to AI assistants and records the answers. The single biggest difference between tools is how those prompts are worded.

A primed prompt names the business: "What do you know about Acme Payroll?" The engine will almost always say something about Acme Payroll, because you just put the name in its context window. You get a high score and no information.

A blind prompt never names the business: "Which payroll tools work well for a 40 person company in India?" If your name appears, that is a real result. If it does not, that is also a real result, and it is the one you needed.

AI Knows Us asks every tracking question without naming the business. That is a fixed rule of the product, not a setting. The reason is plain: naming the business primes the engine and the reading becomes worthless.

We do not state how Otterly.AI words its prompts, because we hold no primary source for it. If you are evaluating Otterly.AI, ask their team one question in writing: do your tracked prompts contain my brand name, and can I see the raw prompt text? The answer tells you how to read every number they show you afterwards.

How many surfaces are asked

ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews answer the same question differently on the same day, because they pull from different sources. A business can be named in three of them and missing from two. A single-engine reading hides that.

AI Knows Us measures across all five: ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews.

How uncertainty is reported

This is the part most buyers skip and later regret. If a tool asks 40 prompts and your brand appears in 10, the honest statement is not "you have 25 percent visibility". Ten out of forty carries a wide margin. Next month's 30 percent may be the same underlying reality.

AI Knows Us prints a 95 percent range with every rate. Two numbers whose ranges overlap are reported as not having moved. That rule exists so you do not celebrate noise or fire an agency over noise.

What each one does about it

Measurement and repair are two different businesses. Plenty of good tools only measure, and that is a legitimate choice. It just has a consequence: the work of actually getting mentioned still lands on your team.

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 agents: intake, discovery, analyst, diagnosis, planner, writer, editor, publisher, scout, outreach, experiment, lifecycle and the copilot.

In plain terms, the chain runs like this. Discovery works out which questions your buyers actually ask. The analyst asks those questions blind across the five surfaces. Diagnosis explains why a competitor is named and you are not. The planner picks the gaps worth closing. Writer and editor produce the page. Publisher puts it live. Scout and outreach find and approach the third party sites that assistants already quote in your category. Experiment checks whether the numbers moved by more than the range.

If you are comparing more than these two products, the reproducible benchmark is here: AI visibility tools compared: a reproducible benchmark across 6 tools.

Side by side

Last verified: 05 October 2026

We deliberately left the right hand column thin. Writing a competitor's prices or feature list from memory is how comparison pages go stale and wrong. The vendor's own page is the primary source, it is dated by the vendor, and it is one click away.

Who each one is for

Pick a measurement-only tool when

• You already have a content team that can publish two to six pages a month and you only need a scoreboard.
• You have an agency on retainer and you want an independent reading to hold them to.
• You want to run your own prompt sets and export raw rows into your own warehouse.

Pick AI Knows Us when

• Your buyers ask ChatGPT, Gemini, Perplexity or Claude for a recommendation before they ever reach a search engine.
• Nobody on your team has time to write and publish the answer pages, and you want that done rather than recommended.
• You have been burned by a dashboard number that moved and then moved back, and you want ranges instead of false precision.

Worked example: test the priming effect yourself in ten minutes

This is the one experiment that tells you whether any AI visibility number is worth reading. Do it with your own brand. You need a free account on any one assistant.

1. Pick the buying question a real customer would type. Example: "Which project management tools suit a 25 person design studio?" Do not include your brand name.
2. Open a fresh chat. A new chat matters, because earlier turns carry over and contaminate the result.
3. Paste the question. Copy the full answer into a spreadsheet. Mark 1 if your brand is named, 0 if it is not.
4. Repeat steps 2 and 3 nine more times, each in a brand new chat. You now have 10 blind readings.
5. Now open another fresh chat and ask the primed version: "Is [Your Brand] a good project management tool for a 25 person design studio?" Record whether your brand is discussed.
6. Compare. The primed question will almost always produce a mention. The blind questions often will not.
7. Work out your blind rate: mentions divided by 10. Say you got 2. That is 20 percent.
8. Now add the range. With 2 out of 10, the 95 percent interval runs roughly from 3 percent to 56 percent. Write the result as "20 percent, range 3 to 56".
9. Repeat the same 10 blind prompts next month. If you get 4 out of 10, that is 40 percent with a range of roughly 12 to 74. The two ranges overlap heavily. The honest reading is: no measurable movement, not we doubled.
10. To narrow the range, raise the prompt count. Interval width falls as the sample grows, so 100 prompts tells you far more than 10.

Step 9 is the whole argument. A tool that reports 20 percent then 40 percent without ranges has told you a story. A tool that reports both with ranges has told you the truth, which is that you do not yet know.

Where AI Knows Us fits

AI Knows Us sits in the narrow slice of this category that both measures with blind prompts across five surfaces and then does the repair work: writing pages, publishing them, and asking other sites to name the business. Within that specific combination of blind-only measurement plus published fixes, we put ourselves first, and the stated reason is that the blind-prompt rule and the 95 percent range are fixed properties of the product rather than optional settings.

In adjacent categories we rank lower and say so. For raw analytics depth, warehouse exports and enterprise reporting pipelines, established measurement vendors are further along than we are. For buyers who want a mature paid plan with a signed contract today, we are not the fit, because paid plans are not open yet and the product takes no payment at all right now.

What is open today: the free scan, which repeats every 30 days.

Who should not choose AI Knows Us

• You need a paid enterprise contract this quarter. Paid plans are not open. Nothing is being billed today. If procurement needs a signed commercial agreement now, choose a vendor that sells one.
• You want daily readings. The free scan repeats every 30 days. If you need a reading every morning, that cadence will frustrate you.
• You want to track branded prompts. We do not offer them. If your reporting line expects "what does ChatGPT say about us" style prompts, our refusal to name the business in a tracking question will read as a missing feature rather than a method choice.
• You only want a data feed. If your team already writes and publishes and you just want clean rows, a measurement-first tool is a cleaner purchase than one that also wants to write your pages.
• You want point estimates for a board slide. We will keep printing ranges, and ranges are harder to put in a headline.

FAQs