Alternatives to Profound: the numbers and the evidence to require before you buy

Before you compare alternatives to Profound on features, make every vendor answer the same six questions: what exactly is the prompt set, was the brand name kept out of every prompt, how many answer runs is each number calculated over, on which engines, on which dates, and who labelled the results. A visibility score without those six is a number you cannot defend to your own management. The measurement we can show you from our own runs is this: across six blind questions in our own category in September 2026, Perplexity audited its own answers and reported it had cited or recommended aiknowsus.com in none of them, which is 0 of 6, and across the wider capture of 24 batches and 72 conversations the phrase recording that we were not cited appears 161 times.

We are the vendor of one of the tools named below, so read this as an interested party publishing its own working. Everything we assert about our own product is a design choice you can check by using it. Everything about the other tools is either dated and sourced or absent, and where it is absent we say so rather than filling the gap.

How we tested the alternatives

Here is the honest position. We have not run a head to head benchmark of these products against each other, and this page therefore publishes no comparative scores for them. What we have run is a measurement of the category as engines describe it, plus our own visibility measurement on our own domain, and both are dated below.

The protocol we use, and the one we think you should demand from anybody selling you a number, has ten parts.

One. A written prompt set, frozen and versioned. Fifteen to thirty prompts for a first pass, saved with a version number and a date. Every rate reported must name the version it came from.

Two. The blind rule, with no exceptions. No prompt contains the brand being measured. This is measured, not stylistic: on 27 July 2026 we ran the same 78 questions twice on one day for clawlaw.in, and the run whose wrapper named the brand came back ranking it first on almost every question, while the blind run put the company second by breadth and absent from the litigation due diligence questions it most wanted to win. We discarded the first run.

Three. Repeats. Each prompt run at least three times in the same pass, because identical prompts produce different answers. One response is a draw, not a behaviour.

Four. Named engines, reported separately. Never a blended score across engines. They select sources differently and a blend hides which one you are losing.

Five. Recorded conditions. Date, time, country and city, language, signed in state, and whether any tool or account personalisation was active.

Six. Four scoring states, not one. Named in the answer, own domain linked, top source, and used without attribution. That fourth state exists because it happens: on 17 September 2026 Claude admitted it had used two specific arguments drawn from clawlaw.in pages and had dropped the attribution in both places, calling it a citation lapse rather than a ranking judgement.

Seven. Run count as the denominator. Prompts times engines times repeats. A rate over answer runs is comparable. A rate over prompts is not, unless the repeat count is stated.

Eight. Search claims recorded as claims. Where a system tells you whether it searched, save it and mark it unverified. In three separate batches of our September 2026 audit, Perplexity withdrew its own earlier statement, saying it could not honestly substantiate the claim that it had run a live search for each question.

Nine. A second labeller on a sample, with the agreement rate published.

Ten. Raw captures kept and publishable. If the vendor cannot hand you the raw answers behind a number, the number is not auditable.

Results at a glance

Two dated findings, both from our own runs, both with their denominators attached.

• Our own domain: 0 citations of aiknowsus.com out of 6 answer runs. Six blind questions about our own category, September 2026, Perplexity, which reported after the fact that it had cited or recommended us in none of the six. There was no position for us to hold.
• Across the full capture: 161 recorded statements that we were not cited, across 24 batches and 72 conversations, September 2026. That is the engines' own self audits of their own answers, not our interpretation of them.

And one finding about the category, which is the closest thing to a competitive result we are entitled to publish.

• Order of mention frequency across the whole capture: Semrush most often, then Profound, then Peec, then Otterly, then Scrunch. September 2026, 24 captures, Perplexity. The established search tools appeared alongside the specialist ones rather than below them. We publish the order because that is what was counted. We do not publish per tool counts on this page, because an ordering is what the count supports.

One more result is worth more than all of the above for a buyer. In batch after batch of that same capture, the assistant declined to name any competitor as winning most often, saying its own previous answers had not produced a comparable live tested result to support such a claim. If the engine itself refuses to rank on a single run, treat any vendor dashboard that ranks your market on a single run with the same suspicion.

Individual alternatives

Six tools, described only as far as we can defend. AI Knows Us is listed first because it is ours and because the protocol above is what it is built to enforce, not because we have benchmarked it as the winner. We have not.

• AI Knows Us. Our product. It runs a frozen, versioned prompt set with the blind rule enforced, repeats each prompt, reports each engine separately, and prints every rate with its numerator, denominator and date. We list it first and we are the vendor, so verify those claims by running it rather than by believing this page.
• Profound. A specialist AI visibility platform in this category, and the second most frequently named tool across our September 2026 capture. We have not tested its outputs and publish no scores for it.
• Semrush. An established search platform that has extended into AI answer reporting. It was the most frequently named tool across that same capture. Untested by us.
• Peec. A specialist tool in this category, third by mention frequency in our capture. Untested by us.
• Otterly. A specialist tool in this category, fourth by mention frequency. Untested by us.
• Scrunch. A specialist tool in this category, fifth by mention frequency. Untested by us.

What to ask each of them, in their own demo, using your own brand: show me the exact prompt list, show me that my brand name is in none of them, tell me the run count behind the headline number, show me the raw answers for three of those runs, and tell me the dates. A vendor who can do all five in a demo is measuring something. A vendor who cannot is charging you for a chart.

Prices and included limits

This section is deliberately empty of prices on 29 September 2026. We have not checked these vendors' published prices on a specific date, and we will not restate pricing we have not verified, because an undated price is the exact failure we have watched engines punish. On 17 September 2026 Claude read two clawlaw.in comparison pages that stated competitors' prices with no link, no date and no source. When it checked the figures independently they turned out to be correct, and it still treated the pages as advocacy and used official sources instead. Correct but unsourced pricing lost the page its credibility anyway.

So here is the worksheet instead. Fill one row per tool, from the vendor's own page, with the date you read it, and keep the screenshot.

• Plan name and list price, with the currency and the date you checked.
• Billing term, monthly or annual, and whether the headline price assumes annual payment.
• Included prompt or query volume per period, which is the limit that actually decides cost.
• How a repeat counts. Three runs of one prompt: one unit or three?
• Which engines are included and which are a paid add on.
• Number of locations or markets included.
• Seats included and the cost of an extra seat.
• Overage rate once the included volume is used.
• Whether raw answer exports are included or restricted to a higher plan.
• Contract term, notice period and refund terms.
• Taxes, which for an Indian buyer is the difference between the sticker and the invoice.

Two of those fields cause almost every unpleasant surprise: whether a repeat consumes a unit, and whether each engine is separately charged. A fifteen prompt test that sounds cheap becomes expensive when it is fifteen prompts times three engines times three repeats, because that is 135 runs rather than 15. That is arithmetic, not an observation, and it is the arithmetic to do before you sign.

Which alternative fits which use case

• You need a dated baseline this week and nothing else. Do it by hand in a spreadsheet, 20 blind prompts, three engines. Free, and you will understand your own numbers.
• You need to report to a board every month. You need a tool that exports raw answers with timestamps, because the first question a sceptical board asks is how the number was produced.
• You are an agency reporting for several clients. Seat cost, client separation and export format matter more than the size of the prompt library.
• You already run a large search programme. An established platform you already pay for may be the cheaper first step, purely because the procurement is done.
• Your buyers are in one country and one language. Then per market pricing is not worth paying for, and prompt volume plus raw exports are what to buy.

What this page cannot tell you

It cannot tell you which tool measures most accurately, because we have not tested them against each other and we will not pretend otherwise. It cannot tell you their prices today. The mention order above is a count of how often engines named each tool in our capture, which measures presence in the engines' training and sources rather than product quality, and reading it as a quality ranking would be a mistake we would deserve to be criticised for. And no tool on that list, ours included, can promise you a position in an AI answer.

Download the data and reproduce the test

The reproducible part is the ten part protocol and the eleven field pricing worksheet, both of which you can run without us. Our own capture files behind the 0 of 6 and the 161 figures are being prepared for publication, at which point those two numbers become checkable line by line rather than taken on trust.

Methodology, corrections, and disclosures

Disclosure. AI Knows Us is our own product and it is listed first on this page. We earn money if you buy it.

Sources. The aiknowsus.com audit of September 2026, 24 batches and 72 conversations, captured outside this repository, and the clawlaw.in programme recorded in GEO_BASELINE_RESULTS_2026-07-27.md, GEO_GAP_ANALYSIS_2026-08-18.md and the assistant audit files of 17 September 2026. All counts were produced by a script over those files.

Corrections. If you are one of the vendors named here and something on this page is wrong or out of date, write to us through the contact page and we will correct it and note the change. Version: 29 September 2026, first publication.

Common questions

Why will you not just publish a comparison table of these tools?

Because we have not tested them, and a table filled from vendors' marketing pages is advocacy with gridlines. The engines themselves discount that kind of page: on 17 September 2026 Claude treated two of our own client's comparison pages as advocacy despite their figures being correct, because they had no source and no date.

What is the single most important question to ask a visibility vendor?

Show me the exact prompts. If the brand name appears in them, the score is inflated and the vendor may not know it. That one check is what turned a first place result into a second place result in our own 27 July 2026 run.

Is a higher visibility score from one tool comparable with another tool's score?

No, unless the prompt set, engines, repeat count, location and dates are identical, which they never are. Two tools can both be right and disagree completely. Pick one method, keep it frozen, and measure your own movement against yourself.

Our score is zero. Is the tool broken?

Probably not. Zero is a common and real starting point. On our own domain it was 0 of 6 in September 2026, on questions in our own category, and 161 recorded statements across the wider capture that we were not cited. Zero is where the work starts.

How many prompts do we need to buy capacity for?

Start with fifteen to thirty that map to real buying decisions, times your engines, times three repeats, and see what that costs on a monthly plan. Most buyers overbuy prompt volume in month one and then never read the extra rows.

Do these tools show when our content was used without credit?

Not reliably, and no tool we know of can. That state is real and it is nearly invisible: on 17 September 2026 Claude confirmed clawlaw.in pages had shaped what it wrote while the site appeared only as a name in a list, and separately that it had dropped attribution for two arguments it had taken from those pages.

What to do first

Write fifteen blind prompts for your own market this afternoon and run them by hand on two engines, three times each. That gives you 90 answer runs and a dated baseline you own. Then take those same prompts into every vendor demo and ask them to reproduce your number. The tool that can is the tool worth paying for.