SiteTell crawls a website for free, names the exact phrase and rule behind every flag, and charges once for the fixes, a cheap first pass at a job most marketing teams never get around to scheduling.
A landing page usually gets written once, published, and then left alone. Nobody schedules a follow-up read-through to check whether it still sounds like a person wrote it. That is the gap SiteTell is built to sit in: a site-wide scan that names the exact phrase, sentence, and rule behind every bit of copy that reads as generic AI writing, then offers a rewrite for it.
The mechanism is more specific than "AI detector." SiteTell splits the job into two parts, and only one of them touches a model at all. Crawling and detection are rule-based: the tool fetches every real page on a site, strips navigation and boilerplate, and flags stock phrases ("delve into," "unlock," "in today's fast-paced world"), sentence-length monotony, and generic openers. Because that half is a ruleset rather than an AI call, SiteTell offers it free and unlimited, whole site, no signup, repeatable as often as a team wants to rescan. The second part, generating a rule-grounded rewrite for each flag, is the part that costs something to run, and it's the part SiteTell charges for.
That split shows up directly in the pricing. The free tier gets a team a full crawl, every flag and score, and a shareable report link. A one-time $20 fee per site unlocks the fix suggestions themselves, plus PDF and CSV export and a markdown or JSON export built for handing to Claude Code or Cursor. Teams that publish often can add an optional $9.99 a month retainer so every new rescan's fixes stay unlocked automatically, rather than paying the one-time fee again each time content changes. None of that is a hypothetical price; it's what the page itself lists, and it means the entire loop, catch the generic copy, understand why it's flagged, and get a way to fix it, costs at most $20 to run once against everything a company has published, or nothing at all if a team is willing to write its own fixes once the flags are shown.
The company also points the tool at itself, which is a useful bit of evidence rather than a marketing flourish. Its own draft homepage scored 6 out of 100 on launch day, mostly from its own "here's what gets flagged" demo copy. After a pass on the pricing text, the shipped page scored 100. That is a small sample of one, but it is at least an honest one, and it's the kind of before-and-after a reader can actually check by visiting the site.
Where this is genuinely useful is the audience a copywriter or content agency rarely gets hired for: a five-person SaaS marketing team that wrote its own landing pages fast, using an AI assistant somewhere in the process, and never had anyone circle back to ask whether the result reads as trustworthy or as one more templated pitch. SiteTell's own framing after Google's February 2026 core update is that generic-sounding copy is now a search risk as well as a conversion one, though that claim comes from the company's own product page rather than an independent report, and is worth treating as their stated rationale rather than a verified ranking effect.
Where it is the wrong tool: anything the ruleset cannot see. A rule that flags "delve into" or a generic opening sentence cannot judge whether a brand's short, punchy, deliberately repetitive style is a real voice or a lazy one; a human editor reading for tone will catch nuance a pattern-matcher won't. It also does not check facts, claims, or legal accuracy, only how the sentence sounds. A team with an actual in-house editor, or one that already runs every page through a real style review before publishing, is likely to find little new here beyond confirmation. And a single-page check, the kind already available as a free browser extension, is enough for anyone who only needs to test one piece of copy rather than track a whole site's drift over time; SiteTell's own FAQ makes that distinction rather than pretending its whole-site crawl is the only option.
The tool's FAQ makes one other point worth sitting with: flagged writing isn't necessarily AI-written, it's just generic, and plenty of the two categories overlap by accident. That is arguably the more useful reframe than the product's own name suggests. The real question a scan like this raises isn't whether a page used AI. It's whether anyone would notice, or care, if it did.