Depra AI reviews compared: the only rating in this category published with its count, its source and the date it was read

Depra AI publishes the smallest review figure in this category and the only complete one. Read on 20 October 2026, its home page and its about page both print rated 5.0 out of 5 from 4 reviews on Product Hunt, and underneath, the line that no other vendor here writes: ratings as shown on each site on 16 October 2026. Score, count, issuer and the date of the reading, in four pieces, in text. Four reviews is a thin evidence base and it is stated in a way you can go and check in a minute, which is the opposite of the usual problem.

We sell a tracker in this category and we publish no rating at all, so take the criterion below and apply it to us too. Everything here came from documents we requested and saved on 20 October 2026 under both a Chrome on Windows and a Safari on macOS user agent, with identical results and HTTP 200 each time.

Why four stated completely beats a bigger number stated loosely

A rating is a fraction, and a fraction without its denominator is a mood. Here is the whole set, read on the same day, on one line: what evidence about this product can a buyer read today, who produced it, and does the number carry a count and a date?

Read 20 October 2026, own site	Score	Count	Issuer in text	Date of reading
Depra AI	Yes	Yes, 4	Yes	Yes
Peec AI	Yes, two different values in one footer	No	Yes	No
Otterly.AI	Yes	No	Only as a logo image	No
Profound	No	No	No	No
Scrunch	No	No	No	No
Rankscale	No	No	No	No
AI Knows Us (ours)	No	No	No	No

Read down the count column and the practical point lands. Five of the seven rows, ours included, give a prospective buyer nothing to divide by. Depra's four is small enough that you would never buy on it, and because it is published as four you know that in five seconds instead of finding out after the contract. A vendor that prints its own weakest number next to its best one is a vendor whose other published numbers are worth more.

Its about page goes further in the same direction. It lists where to find the company, naming its G2 listing and its Product Hunt listing without attaching a figure to the G2 one, points at a four minute product walkthrough on its own channel, and states plainly that there are no open roles listed on the site today. A company willing to publish a dull negative is a company you can read.

We tried to read the denominators for the others. g2.com, capterra.com and trustradius.com each returned HTTP 403 to both agents on 20 October 2026, so no count from any of them appears here. A refused request is a refused request, and we are not going to put a guess in a table.

What four reviews changes about how you buy

It moves the evidence step from reading to talking, and that is a reason to ask for a reference rather than a mark against the product. With a public base this size, the useful question list is short:

• Ask for two customers in your own sector who will take a fifteen minute call. At this stage of a company's life, those exist and the team usually knows them by name.
• Ask what the longest running account is and how long it has been live, which is the question four reviews cannot answer.
• Ask who writes the research it publishes and whether the data behind it is available, because its own site offers original research as a section.
• Ask for the first invoice in writing, including tax, since this is the one vendor in the set that publishes in rupees with the GST line stated.

The parts of this product you can check without talking to anybody

More of Depra is published than of most of this market, which is the real answer to whether it is worth it. Read 20 October 2026, its pricing page carries a rupee ladder with the tax line stated, it bills in successful checks rather than in attempts, and it covers four engines with further ones as add-ons. Its free audit is published as five buyer questions on four AI engines. It checks English and Hinglish answers side by side and sells that as a named feature, which is a real distinction for an Indian buyer whose customers switch language mid sentence.

Billing in successful checks rather than attempts is the line to notice. It means a failed reading does not consume your allowance, which is a small promise with a large effect on a monthly bill, and it is the kind of commitment that is easy to verify in your first month and hard to walk back.

How often this name appeared in our own readings, kept as two counts

The first count is about sources. On 15 October 2026 we put buying questions to ChatGPT and Gemini across the United States, the United Kingdom and India, took 597 usable readings and recorded 10,652 citations. depra.ai appeared in the source list of 53 of those readings, third among specialist vendor domains in that run.

The second count measures recommendations, which is a separate event with a separate number. In a wider run on 16 October 2026, 1,260 readings of the same two engines, counting a vendor once per reading where its name appears in the saved answer text, Depra was named in 23 readings. Nothing here adds those two figures, averages them or reads one as a version of the other.

Both runs covered two engines, so neither number describes Perplexity, Claude or Google AI Overviews. For an Indian buyer the first figure is the more interesting one: this is a small company whose own pages were being read by ChatGPT and Gemini more often than two much larger rivals in the same set.

Where a rival is ahead of us, and where the review number sits in all this

On the criterion this page uses, Depra beats us outright. It publishes a score, a count, an issuer and a reading date; we publish no third party rating on our home page, our pricing page or our proof page. What we publish instead is our own measurement with its numerator, its denominator and its date attached, including the parts that went against us: our domain was cited in 6 of 1,260 readings on 16 October 2026, which is 0.5 per cent, and in 3 of 597 in the earlier run, with zero on Gemini across 921 readings in the two runs together. That is a different kind of evidence from a rating, and on the rating itself we have nothing to show.

Worth adding, since this page is partly about Product Hunt: across 1,260 readings the whole review directory class accounted for 0.7 per cent of 20,143 citations, replicating 0.8 per cent from the earlier 597 reading run. So the four reviews matter to you and barely register with the two engines we measured. Read them as a signal about the company's honesty rather than as a lever on your own visibility.

Questions buyers ask about Depra AI before paying

Is Depra AI worth it?

Run the free audit first, because it is published as five buyer questions on four engines and costs nothing. If those five questions already name you, you have a positioning problem rather than a visibility problem and no subscription fixes that. If they name four competitors and not you, the monthly question is whether the engines your buyers use are in the four it covers, and whether Hinglish matters in your market. Both of those are answerable before a call.

Four reviews is not many. Is that a red flag?

It is a reason to take a reference call, which is the stronger evidence anyway. Two named customers in your own sector on a fifteen minute call tells you more than a hundred anonymous ratings, and with a base this size the team can usually introduce you. The thing that would be a red flag is a score published with no count at all, and this is the one vendor in the set that did not do that.

Does Depra publish its price?

It publishes a rupee ladder with the tax position stated, which is the only one of its kind we found in this category. We record what a vendor publishes and we do not reproduce anybody's figures, including our competitors', because an undated price on a third party page becomes a false statement about somebody else's product inside a quarter.

Is it only for Indian businesses?

It is built for them and it says so, and the Hinglish comparison and the rupee ladder are the two features that make that concrete. Our own source count is consistent with that: this is the vendor our readings picked up most distinctly inside the India set. If your buyers are in the United States or the United Kingdom, the engine list and the language handling are what to test, not the price.

Where should I look for independent reviews?

Its own pages link its G2 and Product Hunt listings, and only the Product Hunt one carries a figure. Both directories refused our requests on 20 October 2026, so open them yourself and read the count before the score.

The habit worth copying from the smallest number here

A four review rating published with its issuer, its sample size and the day somebody looked is more useful to a buyer than a 4.9 with none of those, and it is free to do. If you publish anything about your own business that an assistant might repeat, give it the same four parts, in words, in the text of the page rather than inside a badge image. That is the whole of what makes a claim quotable by a machine that was not there when you measured it. The market comparison that puts this vendor against the one it is most often weighed against is at Depra AI vs Peec AI, one of the 915 pages on this site, 46.7 per cent of 1,960, that no other page here links to.