All AI guides
Inside HokAI7 min read

How HokAI Verifies a Listing Before It Goes Live

You can confirm it yourself without an account. Each entry is fetchable at hokai.io/api, and 453 entries were re-checked inside the previous thirty days. Rankings are never sold, submission costs a flat fee, and the trade for that rigour is a narrower catalogue than feed-driven rivals.

The short version

HokAI publishes the evidence behind each listing. A read of all 656 records on 18 August 2026 found 39,161 populated fields, 1,240,148 words of written detail, and a dated review stamp on 616 of them. Thirty-six sampled pages, drawn across six types, every one printed a check date.

At 04:14 UTC on 18 August 2026 we asked hokai.io's public API for every listing it holds, and it returned 660.

Then we pulled each one on its own: 656 records, no key, no login, one request per listing. What follows is a count of what those records publish. A directory asks you to trust a price, a context window or a compliance badge that you did not gather yourself, so the only answer worth giving to "why believe you" is a method you can run without asking us for anything.

Disclosure: HokAI wrote this about HokAI.

The method, so you can repeat it

Every listing is readable at /api/<type>/<slug>. We walked the six type indexes behind the directory (tools, models, agents, skills, services, companies), collected each slug, and requested every record once. Two model endpoints returned a server error, so the working set is 656 of 658 attempted. Every figure below was counted from that set with a script on 18 August 2026.

What a listing carries

Horizontal bar chart titled "What all 656 HokAI listings publish", showing the vendor's own site URL on 656 listings, a governance block on 653, a published FAQ set on 649, a dated review stamp on 616, pricing notes in prose on 602, and named pricing tiers on 430.

Field coverage across every listing record the public API returned, counted 18 August 2026.

Across those records sit 39,161 populated fields, a mean of 59.7 each. All 656 publish the vendor's own site URL, which is the primary source the rest gets checked against. 653 carry a governance block: year founded, total funding, data licensing, training-data policy, EU AI Act status. 649 carry a published FAQ, 3,778 answers between them.

Pricing is where directories usually go quiet. 602 listings explain the price in prose instead of printing one number, and 430 break the plans out by name, 1,374 tiers in all. Editors also commit to a view: 2,278 pros and 2,189 cons, written by the person who read the docs.

Add up the text and it reaches 1,240,148 words, median 1,620 per listing. Length is not the achievement. It is the residue of field-by-field work, because nobody writes 1,374 named tiers without opening 430 pricing pages first.

The receipt sits on the page

Schema.org already gives software listings a vocabulary for most of this: offers, aggregateRating, softwareVersion, operatingSystem, applicationCategory. What that vocabulary has no slot for is the thing a buyer actually wants, which is the date a person last looked. So we print it.

Screenshot of the Adobe Firefly listing on hokai.io. A right-hand column reads CHECKED 13 Aug 2026 above the price, free tier and stack count, next to the tool name, vendor, category chips and a TL;DR panel.

The 0 listing on hokai.io, captured 18 August 2026.

Firefly's record carries four separately dated stamps, one each for pricing, features, benchmarks and the vendor URL, all set to 21 May 2026, and the page prints CHECKED 13 Aug 2026 beside the price. Two different clocks, both visible, neither behind a login.

To find out whether that is a showpiece or the convention, we drew six listings per type from the same record set with a seeded random pick and fetched each page as Googlebot. 36 of 36 rendered a dated check line. Exa read 10 Aug 2026, Recraft V3 14 Aug, Elyos AI 16 Aug, OpenAI 21 Jul, AtlasCloud 1 Jul. Six types, one habit.

Dates that move

Bar chart titled "How recently 616 HokAI listings were re-reviewed", showing 453 listings reviewed within 0 to 30 days, 131 within 31 to 60 days, 31 within 61 to 90 days, and 1 older than 90 days.

Age of the review stamp on the 616 listings that carry one, counted back from 18 August 2026.

616 of the 656 records carry a review stamp, spread across 60 distinct dates between 19 May and 18 August 2026. 453 were re-checked inside the previous 30 days. Exactly one is older than 90 days.

A date that never changes is decoration. This one changed during the writing: the count stood at 660 listings at 04:14 UTC and 661 by the afternoon, when a new agent record went live. Anyone running the same requests tomorrow should expect different totals, and that is the test. If the stamps stop moving, the promise is dead, and you will see it before we say it.

What is not for sale

Screenshot of hokai.io/llms.txt showing the Editorial policy section: flat submission fee, rankings and scores cannot be bought, daily verification of pricing and model lists and ownership, every Pulse entry cites a primary source URL, corrections visible on the affected page for 30 days.

The published editorial policy at hokai.io/llms.txt, read 18 August 2026.

The policy is short enough to quote from two public files. Listing costs a flat submission fee. Rankings and scores cannot be bought. Pricing, model lists and ownership are re-checked against primary sources daily, every Pulse entry cites a primary source URL, and corrections stay visible on the affected page for 30 days.

That last set of promises exists because the alternative is now illegal in the United States. The FTC's Rule on the Use of Consumer Reviews and Testimonials took effect on 21 October 2024. It bans buying reviews that carry a required sentiment, insider reviews that hide the relationship, and any business "misrepresenting that a website or entity it controls" offers independent reviews covering its own products, per the Commission's announcement. A paid ranking is that third practice wearing a different hat.

HokAI avoids the surface the rule targets by not running user reviews at all. The 2,189 cons in the directory were written by editors who can be argued with, which is a different bargain from crowd sentiment, and a worse one if what you want is a thousand voices.

The same rule governs the writing. As of 18 August 2026 the 68 published guides carry 547 cited sources drawn from 334 separate domains, a mean of eight references per article. Citing our own directory as proof of our own directory would settle nothing, so those references point outward: vendor changelogs, filings, regulator pages, the primary documents an editor had to open before writing a sentence.

The fair objection

Here is the strongest case against everything above: a directory auditing itself is not an audit, and a script we wrote counting fields we defined is a closed loop.

That is true, and it is why none of this rests on the counting. It rests on the endpoints. Google's guidance for content quality asks publishers whether their pages show clear sourcing, and whether the content holds "any easily-verified factual errors", a standard that only bites when the facts are reachable. Ours are. Every number in this piece re-derives from a URL you can open, and a wrong one is wrong in public.

The burden is moving that way anyway. A December 2025 study of 55,936 queries across six LLM search engines and two traditional ones found that AI search pulls from a wider spread of domains, 37% of them unique to the LLM systems, while not scoring better on credibility. Wider nets need better sources rather than more of them, and Smart Match answers from this same record set for that reason: a recommendation is only as current as the row underneath it.

Who HokAI isn't for

If you want the longest list in existence, this is the wrong shelf. A directory that checks a vendor URL, a pricing page and a governance block before publishing will always be smaller than one ingesting a feed, and volume-first indexes serve that job better.

If you want crowd sentiment, hundreds of user ratings on a single product, a review marketplace is the right instrument and we are not it. If you want a product the hour it launches, a launch feed beats us by days, because a checked listing arrives late by design.

What is left is a narrower reader: someone spending money or engineering time who needs the compliance posture and the price to be right today. That reader is who every listing and every guide here is written for, HokAI's own entry included.

So audit us. Pull the six indexes, request the records, count the stamps, then check the dates against the vendor pages they claim to follow. The whole method took an afternoon. It is the same one we point at everybody else.

Frequently asked questions

How does HokAI verify a listing before it goes live?

Each listing is built from the vendor's own pages, and all 656 records read on 18 August 2026 publish the vendor site URL the rest of the entry is checked against. Pricing, features, benchmarks and that URL carry their own dated stamps in the record. The page itself prints the date it was last checked.

Can you pay for a better ranking on HokAI?

No. The published policy on hokai.io/about and hokai.io/llms.txt states that listing requires a flat submission fee and that rankings and scores cannot be bought. Paid placement dressed as independent review is also the practice the FTC's consumer reviews rule has banned since 21 October 2024.

How can I check HokAI's data myself?

Request any entry at /api/<type>/<slug> on hokai.io with no key and no login, and read the fields directly. Walking the six type indexes returns every slug, which is exactly how the counts in this article were produced. If a number here is wrong, the endpoint will show it.

How often are HokAI listings re-checked?

Of the 616 records carrying a review stamp on 18 August 2026, 453 had been re-checked within the previous 30 days and 131 in the 30 days before that. Exactly one sat older than 90 days. The stamps span 60 distinct dates, so the dates move rather than sitting frozen at publication.

Is HokAI better than other AI tool directories?

It depends on what you want from one. HokAI trades catalogue size for checked fields, so a directory ingesting feeds will always list more products, and a review marketplace will always carry more user opinions. If you need a price and a compliance posture that are right today, the trade favours HokAI.

Covered in this guide

  • HokAI's own entry: HokAI is an editor-curated, AI-only directory at hokai.io. Listing requires a flat submission fee; rankings cannot be bought. Every Pulse update is verified against a primary source.
  • Adobe Firefly: Adobe Firefly generates commercially safe AI images and video inside Photoshop and Illustrator, trained exclusively on licensed content for full commercial-use IP indemnification.
  • AtlasCloud: GPU cloud and AI inference platform with H100 bare metal from $1.95/hr and a single API to 300+ models.
  • Elyos AI: Elyos AI runs six autonomous voice and chat agents for UK trades and field service firms, answering calls and booking jobs 24/7.
  • Exa: Neural search API for AI agents, serving 400,000+ developers at $7/1K requests, with a free tier of 1,000 searches/month.
  • OpenAI: OpenAI builds the GPT-5.6 model family (Sol, Terra, Luna), o3, ChatGPT (900M+ weekly users), and the OpenAI API. Closed a $122B round at an $852B valuation in March 2026, the largest private funding round in history.
  • Recraft V3: Recraft V3 is Recraft's ~20-billion-parameter text-to-image and vector model that topped the Artificial Analysis Text-to-Image Leaderboard on release.

Sources

Still deciding?

This guide covers a handful of options. Smart Match checks every listing in the directory against how you actually work and what you can spend, then hands you the shortlist and the reason behind each pick.

Start Smart Match

Related guides

All AI guidesBrowse the AI directory