Methodology
Traditional directories wait for submissions and a human to categorise them. Verdy's is maintained by agents: they find products as they launch, read each one, check what they read, score it, and file it.
Twice a day Verdy reads the front pages of TinyLaunch and IndieLaunch, and category searches for incumbents. Sources whose terms forbid a commercial directory from reading them are not used. Only the product's own address is kept; listing-page links and tracking tags are stripped, and a product already listed is not queued twice.
Verdy reads every site as VerdyBot, never disguised as a browser, and skips any site whose robots.txt blocks it — launch sites and products alike.
An agent reads the product's homepage and, where the crawler found them, its pricing and docs pages — nothing else. It returns what the product does, who it is for, the problem it solves, a category from a fixed list, pricing, whether there is an API, and answers to six questions an assistant needs answered before recommending software.
Each fact must come with a quote copied from the site. Verdy then checks, in code, that the quote is really on the page. If it is not, the fact is recorded as not stated. A starting price the quote does not contain, or a free plan it never mentions, is dropped even when the quote itself is real. That is why some listings say “pricing not stated”: the site did not say, and Verdy will not guess.
The Verdy Score is a weighted average of six parts. A part that could not be measured is left out of the average rather than counted as zero. Each product page shows every check behind its score.
| Part | Weight | What it measures |
|---|---|---|
| SEO | 20% | The website audit score: a crawl of up to twelve pages, checked against the fundamentals search engines and assistants rely on. |
| Performance | 15% | Google PageSpeed's mobile performance score for the homepage. Left out of the average when PageSpeed could not run. |
| Accessibility | 10% | Google PageSpeed's accessibility score for the homepage. |
| Product clarity | 20% | Whether the site says, in its own words, what the product does, who it is for and what it costs — plus a main heading and a meta description. |
| Trust | 15% | HTTPS, a privacy policy, terms, an about or contact page, linked social profiles, and whether it is a live product rather than a placeholder. |
| Agent readability | 20% | Half plumbing — llms.txt, AI crawlers allowed, structured data, a sitemap, pricing readable without JavaScript. Half a test: could an agent answer six questions about the product from the site alone, with a quote to show for each? |
Category pages, alternatives and “new this week” are generated from the evaluated products, not written by hand. Pages with too few products are kept out of search indexes until they are worth landing on. Listings are re-read when the scoring changes, and products whose site stops responding are retried before they drop out.
The same catalogue is queryable. GET https://www.verdy.site/api/v1/search?q=… returns ranked products with the reasons for each, built from verified fields; GET https://www.verdy.site/api/v1/tools/{host} returns one product. “free” and “api” in a query are treated as requirements, not keywords. See llms.txt.
Claim it from its page with a Google sign-in, for free. If a fact is wrong, the fix is on your site: say it clearly there, and the next read picks it up. To leave the directory, block VerdyBot in robots.txt or ask for removal.