Methodology

Finding new sites is easy; knowing which numbers to trust is the hard part. This page lists every source we use, the rules a site must pass to be ranked, and what each badge means — so you can decide how much weight to give a number.

Data sources

• TabAPI — monthly visits and up to 12 months of history, traffic channels, top countries, top keywords, engagement (bounce rate, pages per visit, visit duration), global rank, and backlink / domain-rating data. It is our primary traffic source. We also use its WHOIS lookup when RDAP has no record.
• DataForSEO — Labs estimated organic traffic (ETV) for the United States, used as the second source for keyword-led sites; and search results for model and use-case keywords, used to discover candidate domains.
• Tranco — a public research ranking of the top one million domains, aggregated from several providers. Used to check that a site's traffic is the right order of magnitude.
• RDAP — the registries' official registration data, for the domain's registration date. Each site page links to the record.
• Certificate Transparency logs — first TLS certificate issued for the domain, a supplementary discovery source and an extra launch-date signal where available.
• Our own crawl — each site's homepage and pricing page: HTTP status, tech stack, analytics tools, on-page SEO, payment-provider fingerprints and listed prices.
• A large language model, run in our offline batch pipeline — classifies whether a site is an AI tool and what kind, writes the one-line description and the "AI take" notes, and reads prices off pricing pages. It never produces traffic or revenue numbers. Each site page names the model and the month the analysis was generated.

Which sites are included

A site appears in the rankings only when all of these hold:

1. The domain was registered on or after January 1, 2025 (RDAP; WHOIS as a fallback). No registration date, no listing. Domains we first see on a newly-registered-domains list are re-checked against RDAP before they can be listed; until then their profile shows “First seen in newly-registered list”, which may be 1–3 days after the actual registration.
2. It is an AI tool — not a news site, directory, template or landing-page shell. Directories are tracked but not ranked.
3. Its homepage returns a 2xx status after redirects. Sites that go down are removed from the boards; their pages stay, marked.
4. It is not an adult, scam or otherwise illegal site. Sites flagged this way are never listed.
5. It has at least 3 months of traffic history and at least 5,000 visits in the latest month.
6. Its cross-check verdict is not unverified (see below).

Thresholds

Status: approved on 2026-10-01. The values come from a 50-site accuracy test run in October 2026 and are read directly from the configuration the scoring job uses.

Cross-checking and evidence badges

DataForSEO's estimate covers US organic search only, so we never compare it with total visits. We compare it with visits × organic share × US share from TabAPI. That only makes sense for keyword-led sites; brand-led and all-in-one sites get most of their traffic from direct visits and brand searches, so for them we check the order of magnitude against the Tranco rank instead. Each site gets one verdict:

verified

Keyword-led site: the two traffic estimates are within ±10× of each other.

rank_checked

Traffic magnitude agrees with the independent Tranco rank.

single-source

Only one source covers the site so far — typically a new site not yet in the Tranco top million and without a DataForSEO estimate. It is ranked, but you are looking at one provider's estimate.

rank_mismatch

Brand-led site whose traffic and Tranco rank disagree. Flagged in the site's red flags for review.

unverified

The sources disagree beyond the band. Not ranked; the site page shows both values.

Revenue estimates and A/B/C grades

We don't know any site's revenue. We model a monthly range from traffic and the site's own pricing, and grade how much evidence is behind it:

• Subscriptions: visits × sign-up rate (0.5%–3%) × paid conversion (2%–8%) × monthly price (lowest tier for the low end, the middle tier for the high end).
• One-time purchases: visits × purchase rate (0.1%–0.5%) × price.
• Ads: visits × pages per visit × RPM ($1–$8 per 1,000 page views).

• Grade A — grade B evidence plus a publicly self-reported revenue figure from a comparable site in the same category as an anchor. No anchors are configured yet, so no site is currently grade A.
• Grade B — a payment provider (Stripe, Paddle, Lemon Squeezy, PayPal, Creem, Dodo, Gumroad …) found in the site's own pages — a checkout link, payment script or first-party statement, not a passing mention — and real prices read from its pricing page.
• Grade C — traffic only. Shown on the site page for context, never ranked on the revenue board.

Update frequency

• Traffic: monthly. The new month is fetched between the 8th and 12th, and the ranking snapshot is written on the 13th.
• Cross-check (DataForSEO and Tranco): monthly, before the snapshot.
• Homepage and pricing crawl: rolling, about one-thirtieth of the sites each day.
• New-site discovery: daily.
• AI model releases: every 30 minutes.

Past snapshots stay available in the monthly archive.

Known limitations

• About a one-month lag. TabAPI publishes a month's traffic roughly a month later, so the current snapshot describes the previous full month.
• ETV lags behind fast growth. DataForSEO's organic estimate updates slowly, so a site that just took off can look inconsistent. That is why a matching Tranco rank can clear a keyword-led site.
• Single-source sites. Many young sites are covered by only one provider. They are ranked with a single-source badge; treat their numbers as one estimate, not a confirmed figure.
• Site type is inferred. Whether a site is brand-led or keyword-led decides which check applies. In our 50-site test the classifier matched manual labels 86% of the time; mistakes tend towards the more lenient check.
• Payment detection is conservative. We count only strong fingerprints, so sites that load their checkout entirely inside an app may show no provider and stay off the revenue board.
• All numbers are estimates. Nothing here is data reported by the sites themselves.

Corrections and removal

If you run a listed site and want it removed, or you think a number is wrong, email hello@newsiteradar.com from an address at the domain (or tell us how to verify you control it). We remove verified sites from the boards and their site page.