katman / AI visibility tools
FIRST-PARTY COMPARISON · 11 PRODUCTS · CRAWLED 4 SEPTEMBER 2026
AI visibility tools, compared
Eleven products that measure whether AI answer engines name your brand, with what each costs and what each cannot do. Every price and capability below was taken from the vendor's own pages on 4 September 2026, not from their marketing claims about each other. We are in the table, in alphabetical position, with the rows we lose left in.
The field
| Product | Entry price | What you get at that price | Strongest at | Does not do |
|---|---|---|---|---|
| AthenaHQ | $295/mo | 10 models, 4 brands | Executive dashboards and PR monitoring | No outcome verification |
| Brandlight | Not public | Enterprise, SOC 2 | Enterprise procurement and support | No public pricing, no prompt tracking shown |
| Evertune | $800/mo | Thousands of prompts | Volume sampling and ad integration | No share of voice; the most expensive entry point |
| katman us | $299 once | 5 model families, 3 months of re-runs | Outcome verification, language splits, technical audit, PDF | No sentiment scoring, no daily tracking, no agency portfolio |
| Knowatoa | $59/mo | — | Cheapest entry, content drafting | No competitor extraction, no outcome verification |
| Peec AI | ~$29/mo | 25–50 prompts, 3 models | Cheapest per prompt, PDF export, Looker | No competitor extraction, no citation analysis |
| Profound | $99/mo | 50 prompts | Agent analytics, deepest prompt tooling | No scheduled white-label reports |
| Rankscale | $99/mo | 50 prompts | Widest feature spread at the price | No crawler log monitoring |
| Scrunch AI | $250/mo | 1 brand, 3 languages, 2 competitors | Enterprise scale and services | No outcome verification; two competitors at entry |
| Trakkr | £79/mo | 1 brand, 50 prompts, 8 models, 1 seat | Evidence-first design, 8 models, actions with fixed measurement plans | No sentiment; no language splits shown |
| xFunnel | Free tier | — | Free entry, experiments platform, GA4 | No competitor extraction shown |
Otterly returned 403 to our crawler and Goodie and Bluefish served nothing crawlable, so all three are excluded rather than guessed at.
What the category has settled on, and what it has not
Twenty capabilities, matched across ten vendors on the phrases a product page would actually use. The middle column is how many of the ten show evidence of it — which is the more useful number, because it tells you what is table stakes and what is still contested.
| Capability | Vendors with it | katman |
|---|---|---|
| Topic or audience segmentation | 10 of 10 | No. Questions carry an intent, but there is no filter yet. |
| Sentiment or perception scoring | 10 of 10 | No, and not planned. Scoring how a model feels about you is softer than what it answers. |
| Multi-model panel | 10 of 10 | Yes — OpenAI, Anthropic, Google, xAI, Perplexity. |
| Agency or multi-client portfolio | 10 of 10 | No. |
| Share of voice | 9 of 10 | Yes, with the denominator printed beside it. |
| Scheduled daily prompt tracking | 9 of 10 | No. Re-runs are on demand, up to twice a week for three months. |
| CSV, API or MCP export | 9 of 10 | Not for customers yet. PDF and a shareable link only. |
| Answer accuracy checking | 9 of 10 | No. |
| AI crawler log monitoring | 9 of 10 | Partial. A browser beacon cannot see crawlers; that needs server logs. |
| Citation and source analysis | 8 of 10 | Yes — the source graph, with which model families cited each domain. |
| Content briefs | 7 of 10 | Yes. Briefs, not generated articles. |
| Competitors extracted from answers | 7 of 10 | Yes, counted from the stored answers rather than estimated. |
| Recommended actions | 6 of 10 | Yes, each with a measurement plan fixed before the work starts. |
| Alerts | 6 of 10 | No. |
| Scheduled white-label reports | 5 of 10 | No. Reports are sent manually into the client's panel. |
| Outcome verification | 5 of 10 | Yes, and it refuses to report a change across a changed experiment. |
| Geography and language splits | 5 of 10 | Yes. Questions are asked in every language you sell in. |
| Technical readiness audit | 4 of 10 | Yes, free and public. |
| PDF export | 2 of 10 | Yes. |
Read the bottom of that table rather than the top. Segmentation, sentiment and multi-model panels are table stakes — everyone has them, so nobody wins on them. The contested rows are the ones that decide whether a tool can answer did the work change anything: five of ten verify outcomes, five of ten split by language, and two of ten hand you a file.
How to choose, in one paragraph
If you want to watch many prompts cheaply every day, the price ladder starts at Peec and Knowatoa and runs through Profound and Rankscale. If you are buying for an enterprise with procurement and a security review, Brandlight, Scrunch and Evertune are built for that and priced for it. If you want one measurement you can act on, hand to someone else, and check again afterwards, that is a smaller field: Trakkr and katman. And if you have not measured anything yet, start with a free audit — ours or anyone's — because the first question is whether AI crawlers can read your site at all, and that answer costs nothing.
What none of them can do
No product in this category can guarantee that a model will recommend you, and every honest one says so. Model answers change by day, country, account and search provider. A tool that reports a number without saying which questions produced it, in which language, on which date, is reporting a mood.
Two failures are common enough to check for before you buy. Branded questions inside the score: a model handed your brand name almost always recognises it, so any tool that counts "what is [your brand]" toward visibility is flattering you. One language reported as global: a brand can be absent in one market and recommended in another, and averaging that is a category error. Ask any vendor for both answers in writing.
Questions
What is the best AI visibility tool?
There isn't one, and any page that names a single winner is selling it. The twelve products in this category split cleanly by what you are trying to do: track many prompts daily and cheaply (Peec, Knowatoa, Profound), buy enterprise scale and support (Brandlight, Scrunch, Evertune), or get one measurement you can act on and verify (Trakkr, katman). Pick the row, not the ranking.
How much do AI visibility tools cost?
$29 to $800 a month as of 4 September 2026, priced by how many prompts you track. Peec starts around $29, Knowatoa $59, Profound and Rankscale $99, Trakkr £79, Scrunch $250, AthenaHQ $295, Evertune $800. Brandlight and Otterly do not publish prices. katman is the outlier at $299 paid once.
Why does katman charge once when everyone else charges monthly?
Because a subscription to a dashboard has to be visited monthly to be worth anything, and we do not sell a dashboard. The $299 includes three months of re-runs at up to two a week, so you can re-measure the moment you ship something rather than waiting for a billing cycle. The recurring charge, when there is one, is for us doing work — extending your question set and writing briefs — because that is the only part that genuinely recurs.
What do most of these tools not do?
Half of them cannot tell you whether the work you did changed anything: only five of ten verify outcomes. Half do not split results by language, so a brand absent in one market and recommended in another reads as a single confident number. Two of ten give you a file you can keep.
Do any of them guarantee that ChatGPT will recommend my brand?
None of them, and you should walk away from any that do. Nobody sells placement in an organic AI answer. What is sellable is measurement, evidence and work.
How was this comparison made?
Every vendor's own sitemap was crawled on 4 September 2026 — up to fourteen pages each, plus the full public documentation of the closest competitor — and each capability was matched against the phrases a product page would actually use. Prices come from public pricing pages. Nothing here was verified by buying and using the products, and vendor claims about their own products are recorded as claims.
Is katman in its own comparison?
Yes, in the same table with the same columns, including the three rows we lose. A comparison written by a vendor that puts itself first is an advertisement with a table in it.
Method and limits
Each vendor's sitemap was fetched on 4 September 2026 and up to fourteen pages per site were crawled, plus the complete public documentation of the closest competitor. Capabilities were matched by phrase, so a product that has a feature and never describes it on a public page will read as missing here. Prices are entry-tier public prices excluding tax; several vendors publish none. Nothing in this comparison was verified by buying and using the products, and every claim a vendor makes about its own product is recorded as a claim. If you run one of these products and something above is wrong, tell us and it will be corrected with the date of the correction attached.
Measure first, choose second. The free audit checks whether AI crawlers can reach and read your site, which decides whether any of these products can help you at all.