AI Knows Us vs Semrush AI Toolkit: how the two measure AI visibility, and what each does next

The problem this page is about

A buyer types a question into ChatGPT, Gemini, Perplexity or Claude. The answer names three companies. Yours is not one of them. Nobody clicked a search result, so nothing showed up in your analytics. There is no impression count, no position, no referrer. You only find out when a sales call opens with "I asked ChatGPT and it suggested someone else."

So you go looking for a tool. Two kinds exist. The first kind adds AI questions to a platform you already use for SEO. The second kind does nothing but AI answers. Semrush AI Toolkit is the first kind. AI Knows Us is the second kind. The rest of this page sets out what each one measures, what each one does about it, and who each one suits.

What each one measures

Semrush AI Toolkit

Semrush describes the AI Toolkit on its own product page as a way to see how a brand shows up in AI search and AI chat tools, including the prompts that trigger mentions and the sentiment of those mentions. Read the current feature list on the source page, because platform features change often: semrush.com/ai-toolkit.

The structural point, and the reason people compare the two, is placement. The AI Toolkit sits inside Semrush. If your team already lives in Semrush for keyword research, site audits and backlink work, AI visibility becomes one more tab rather than one more login and one more invoice.

AI Knows Us

AI Knows Us measures across five surfaces: ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. Three design choices shape what the numbers mean.

• Blind questions. Every tracking question is asked without naming the business. Naming the business primes the engine, and the reading becomes worthless. If you ask "what do you know about Acme Logistics?" the model will talk about Acme Logistics. That tells you nothing about whether it would have raised Acme on its own.
• Five apps, not one. The five apps answer the same question differently on the same day, from different sources. A reading from one app is a reading about one app.
• A range with every rate. Every rate is printed with a 95 percent range. When two numbers have ranges that overlap, the report says the number has not moved. That rule stops a handful of questions being read as a clean percentage.

What each one does about it

Measurement is the easy half. The hard half is changing what the models say, which means changing what is published on the open web about you and what other sites say about you.

Semrush is a large platform with a wide feature surface across SEO, content and advertising research. What the AI Toolkit itself recommends or automates is set out on its product page; check it there rather than taking a competitor's word for it.

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. The work is split across 13 agents, and this is the clearest checkable fact on this page. As of 05 October 2026 the named agents are:

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. the copilot

That is thirteen named roles, not a marketing round number. Four of them (writer, editor, publisher, outreach) exist because a dashboard that only reports a drop leaves the actual job undone.

Side by side

Where a cell says "see Semrush's product page", that is deliberate. We will not paraphrase a competitor's feature depth from memory and present it as a fact.

Who each one is for

Choose a platform feature set like Semrush AI Toolkit if

• You already pay for the platform and want one login.
• Classic SEO is still the main channel and AI answers are a second question.
• You have an in-house team or agency already producing and publishing content, so you need data and not hands.
• You need the AI numbers to sit beside keyword and backlink numbers in one report for a board or a client.

Choose a standalone blind-prompt tool like AI Knows Us if

• Your buyers ask an AI app for a recommendation before they ever open a search engine.
• You care whether the model raises you unprompted, not whether it can describe you when asked.
• You want one number per surface rather than a blended score you cannot take apart.
• Nobody on the team has time to write and publish the pages that would change the answer.

Run the test yourself: eight numbered steps

You do not need either product to get a first reading. You need an hour and a spreadsheet. This is the same shape of test AI Knows Us runs, written out so you can repeat it.

1. Write down the question a buyer would actually type. Not your brand name. For example: "best freight forwarder in Chennai for small exporters".
2. Make 20 variants of that question. Change the city, the buyer size, the use case, the wording. Keep your business name out of all 20.
3. Open a fresh session in each app you care about. Logged out or in a temporary chat, so past conversations do not colour the answer.
4. Paste each question once. Do not follow up. Do not nudge. A nudge is priming.
5. Record one of two values per question: mentioned, or not mentioned. Ignore position for the first pass.
6. Count the mentions. Divide by 20. That is your raw rate.
7. Put a range around the rate. With 20 questions the range is wide, and that is the honest finding. Use a Wilson score interval at 95 percent.
8. Repeat on the same weekday a month later, with the same 20 questions, and compare ranges rather than points.

Worked example you can copy

September reading: your business was mentioned in 7 of 20 blind questions on one app. Raw rate 35 percent. The 95 percent Wilson range for 7 of 20 runs roughly from 15 percent to 57 percent.

October reading: 9 of 20. Raw rate 45 percent. The 95 percent Wilson range runs roughly from 24 percent to 67 percent.

The headline temptation is "visibility up 10 points". The ranges overlap heavily, so the correct report is: no detected change. To call a 10 point move you need far more questions, or a bigger gap. This is exactly the rule AI Knows Us applies: two numbers whose ranges overlap are reported as not having moved. If a tool shows you 35 percent and then 45 percent with no range at all, you cannot tell a real gain from sampling noise.

For a longer version of this method applied to one engine, with the prompt list and the scoring sheet, see #.

Where AI Knows Us fits

In the narrow category of blind-prompt AI visibility measurement that also produces and publishes the fix, we put AI Knows Us first, and here is the stated reason: the same system that takes the reading also writes the pages, publishes them and runs outreach, through the writer, editor, publisher and outreach agents listed above. That removes the gap between a report and a change. It is a reason about scope, not about quality, and you should test it rather than take it.

In the wider category of search marketing platforms, we rank honestly below Semrush. AI Knows Us does not do keyword research, site audits, backlink analysis, rank tracking, ad research or competitor traffic estimates. Semrush does. If that broader job is the job, the platform wins on coverage and this is not a close call.

Other tools in the AI visibility category worth your shortlist, named without claims about their features: Profound, Peec AI, Otterly.AI and Scrunch AI. Read their own pages for pricing and coverage.

AI Knows Us is made by Clyra Labs in India.

Who should not choose AI Knows Us

• Teams whose main channel is still classic organic search. If most revenue comes from blue links, buy the tool that tracks blue links. AI answers can be a quarterly check.
• Teams that need procurement-grade vendor paperwork today. Paid plans are not open, which means there is no invoice, no contract tier and no procurement trail to show finance.
• Teams that want a dashboard only. If you have writers and an editorial calendar already, a measurement-only tool inside your existing platform may be the cheaper fit.
• Teams that want a single blended score. Five surfaces with five separate rates and five ranges is more numbers, not fewer. Some reporting cultures do not want that.
• Anyone who needs a . No tool can promise that. A model's answer is not a slot you can buy here.

FAQs

Sources

• Semrush AI Toolkit product page
• Semrush pricing page
• Profound, Peec AI, Otterly.AI, Scrunch AI