Blog · 24 September 2026 · By HeardOf

Eight questions to ask an AI visibility tool before you buy, and what nineteen tools' own pages say to each

Where do these eight questions come from?

From three reads of the same nineteen AI visibility tools' own pages: pricing pages served on 23 September 2026 and rendered on 24 September, and method, product, docs and help pages on 24 September. This post adds no count; every figure is copied from the live post it links, with the read date that post prints. What it adds is the order.

If you are working out how to choose an AI visibility tool, a vendor's pages answer some of these and not others; a negative means not stated on the pages read, which each linked post lists. Our own answers are in our post on which engines we check, as updated 23 September 2026, and are not set beside the counts here; they come from our own systems and you cannot check them.

1. Can you read the answer the engine gave, and the URLs it cited?

Sixteen of nineteen say on their own pages, read on 24 September 2026, that you can open the answer text: fifteen in so many words — full, exact, actual, complete, unabridged or raw response — and one, SE Ranking, as responses you can review; three say nothing on the pages read, Evertune, Goodie and Bluefish. Eleven of nineteen say the cited URLs or sources are shown with the answer they belong to; seven describe citations only as domains or pages ranked across all answers; one, Bluefish, says nothing. Both counts are in can you see the answer, and who got named instead.

A visibility score is one step removed from what you want to see: a rate, over prompts, of answers that named you. The line between the eleven and the seven is per answer against aggregate: Peec's docs put a "Sources sidebar: All URLs the AI model referenced or cited when creating the response" on every chat; Evertune's FAQ says "Content Analytics reports every domain and URL AI models return as a source", a list across answers.

2. Will it name the brand you did not add?

Eleven of nineteen say their product lists or surfaces brands you did not add, as read on 24 September 2026; five say they show which brands were mentioned, or a competitor set found automatically, without saying whether the names come from the answers or must be declared first; three say nothing on the pages read. Same post as question 1.

The brand taking your answers may not be one you typed in. Otterly's help centre, in an article dated 21 May 2026 and so labelled historical, says its More Detected Brands view "shows all brands that have already appeared in AI answers for your tracked prompts — even ones you haven't added as competitors yet"; SE Ranking's tracker page takes "up to five competitors", which does not say whether a sixth is listed. For a vendor among the five, ask: do the names come out of the answers, or in from your settings?

3. Does it read the chat window or call an API?

Thirteen of nineteen say, on their own pages as read on 24 September 2026, how they collect their data: eight through the chat interface, four through both an API and the interface, and one through an API. Six say nothing on the pages read. The count and every quote are in what nineteen of them say about how they get their answers.

The route is part of what the figure means: Peec's docs say an API answer can differ from "what users see in the actual interface", as read on 24 September 2026 for HeardOf vs Peec AI. Profound's features page: "Other tools pull AI responses from the API. We capture directly from the browser." Conductor's docs: "We use official APIs provided by the AI search engines we support to get the content from responses".

4. Which model version produced the number?

Six of nineteen name a version string somewhere on their own pages, read on 24 September 2026; thirteen do not. The methods post counts a version string only: "the latest model" and "Sonnet" do not count.

Naming one is not naming all: Ahrefs names Claude Sonnet 4.6 and calls everything else "the default model", and Peec names versions only for its paid API models, since for the scraped engines "the platform decides which model to use". Semrush's knowledge base says its toolkit "analyzes responses generated by the latest model of ChatGPT in search mode", a mode and a product but not a version, and none of its nine pages read on 23 September 2026 for Semrush AI Visibility Toolkit vs HeardOf prints one; Surfer says it does not pick a model at all. Neither is counted.

5. How many times a day does it ask each prompt?

Two say once a day in so many words, Profound and Peec, as read on 24 September 2026; three use wording or an example that amounts to one run per prompt per engine per day, Otterly, AmICited and SE Visible; five give a schedule, Scrunch, Rankscale, Ahrefs, Athena and Conductor; Evertune gives a sample count with no cadence, 100 per prompt per model; Surfer says multiple queries per day with no number; five say daily with no count, Goodie, Semrush, HubSpot, Writesonic and Similarweb; two say nothing on the pages read, Bluefish and BrightEdge. The breakdown is the methods post's, and it sums to nineteen.

The methods post puts the case for more than one answer: it and our post on which engines we check, as updated 23 September 2026, both cite Schulte, Bleeker and Kaufmann (arXiv, 8 April 2026), who put the standard error of a single run at 0.370 and recommend at least seven runs per prompt per day for brand-visibility monitoring. A vendor whose pages state no cadence leaves a buyer unable to tell what its dashboard is a sample of, as the methods post puts it.

6. What does it count when nothing comes back?

One of nineteen states a rule that names AI Overviews and says how a prompt without one counts, as read on 24 September 2026: Otterly, whose KPI definitions dated 4 September 2026 say "When AIO are not triggered, they do not count towards that formula." None of nineteen states, on the pages read, how a failed fetch enters its figures, whether it is dropped, retried, counted as an answer with nobody in it, or left as a gap. Both counts are in the methods post; the Otterly rule is also quoted, as read on 23 September 2026, in HeardOf vs Otterly.

An answer that names nobody is a third case, not the same as either. This is the denominator. Profound's metrics documentation states a general rule, "Runs in which no brand was mentioned are excluded from both the numerator and the denominator", without naming AI Overviews. The worked example in HeardOf vs Profound shows the size of it: ten made-up answers, four naming you, three naming only a rival, three naming nobody, come out at 4 of 7, 57%, under a rule that drops the empty ones and 4 of 10, 40%, under a rule that keeps them. Neither is wrong; they are different fractions.

7. Do its own pages agree on how many engines it checks?

Six of nineteen give a number on one page that another page, or the same page, contradicts, as read on 24 September 2026; three more give no number and a different list on each page; ten are consistent or give a single list. The methods post names each.

Otterly says "six AI platforms" in its home-page FAQ and "seven AI engines" on its features page. Goodie's pricing page says "Up to 13 engines" in one tier and "all 12 engines" in the FAQ below it. Two pages written at different times can both be honest and still disagree; ask which page is current before the count goes into a spreadsheet.

8. What does it cost, and does its page show it?

Fifteen of nineteen show a plan price on their own pricing page, served on 23 September 2026 or drawn in a browser on 24 September; four do not, Profound, Bluefish, Conductor and BrightEdge, and the third-party estimates that circulate for three of them are printed as the third party's. The count, each page's entry tier and the ten things that could not be verified are in what nineteen AI-visibility tools charge.

Read the page in a browser, and click every tab. Peec's page serves no figure to a script and, rendered in headless Chrome on 24 September 2026, shows Starter at $95 a month, Pro at $245 and Advanced at $495 on monthly billing; the pricing post went out on 23 September counting Peec among the vendors that show no price. Profound's page, drawn on 24 September 2026, opens on Trial, Free, and Enterprise, Custom, and draws "$399/month" for an added client workspace only after a click on its For agencies tab, per HeardOf vs Profound. Among thirty B2B software pricing pages read the same day, none of the twenty-seven we could read drew a plan price only with JavaScript, per 30 B2B software pricing pages.

Why not take the answers from a vendor's best-of list?

Because of where the author sits. Fourteen best-of and top-N lists of AI visibility tools on the Profound, Otterly, AthenaHQ and Geoptie blogs name two or more tools, and all fourteen put the author's own tool at position one, rendered in a browser on 24 September 2026; none of Peec's 44 posts matches the rule. Of the 179 prices those lists print for other tracked tools, 30 are not on that tool's pricing page as rendered the same day, per the vendors' own best-of lists, found in a 574-post census of the five blogs read on 23 September 2026. Take the eight questions to the vendor's own pages, and to the vendor.

What are the eight, in one table?

Eight rows, one per question, each count as the linked post states it.

How was this read?

Nothing here was measured for this post. The eleven live posts linked above were fetched on 24 September 2026 between 18:35:31 and 18:35:35 UTC, and every count and quotation was copied from their live text, the text as corrected that day; the read behind each count is dated in the post that carries it and in the table above. Every negative is scoped to the pages those posts list; a page older than 90 days on its read day is labelled historical where quoted.

Everything this post says about our own system comes from our post on which engines we check, as updated 23 September 2026, and carries that provenance in the sentence that says it; you cannot check it. Everything else carries a post, a page and a day.

Semrush AI Visibility Toolkit vs HeardOf: what each one measures, and what Semrush does better