Cloudflare will grade Verified bots on behavior and publish trust scores on Radar

Verified bots will now be judged on behavior, not just identity

Cloudflare has begun scoring the bots in its Verified bot program on their ongoing conduct, rather than treating a confirmed identity as the end of the assessment. According to the company's blog, the system watches whether verified crawlers keep honoring website preferences over time, produces an explainable "trust band" for each bot, and will eventually let customers set a threshold that decides how that traffic is handled.

Verification solved half the problem

Cloudflare requires Verified bots to meet three criteria: identify themselves accurately and securely, do only what they say they will do, and refrain from abusive behavior. Bots prove who they are by signing requests with Web Bot Auth, and BotBase catalogs those identities in an open format.

In Cloudflare's framing, identification is largely handled; what has been missing is continuous evidence that a known bot behaves once it is let in. The company distinguishes between risk — the chance that a single request is harmful, which can shift from moment to moment — and trust, which builds through repeated, consistent behavior.

What Cloudflare measures

The evaluation starts with two signals.

The first is whether a bot respects published instructions. The system compares a crawler's observed traffic against a site's robots.txt over time, consistent with RFC 9309. Isolated violations carry little weight; Cloudflare says it looks for repeated patterns that disregard stated preferences. The company also notes that robots.txt sets a floor rather than a ceiling — anything the file does not mention is not automatically permitted, and sites can communicate that continued access is unwelcome through other signals.

The second is whether a bot responds to direct feedback. A site under strain can reply with HTTP 429, "Too Many Requests", a directive to ease off, sometimes with guidance on how long to wait. Cloudflare checks whether the bot then slows down, pauses, or keeps requesting at the same pace. Consistently backing off raises trust; repeatedly ignoring the signal lowers it.

The output is a trust band: a range reflecting how consistently a bot respects site preferences, along with the observed behaviors behind the assessment. Cloudflare stresses that this is not a permanent judgment about a bot or its operator — a meaningful caveat, since the company notes that problematic behavior often stems from a misconfiguration or a bug rather than malice.

The precedent: Perplexity's delisting

Cloudflare points to its 2025 delisting of Perplexity's bot as the kind of failure the system is meant to catch. The bot, according to the company, sought access to resources on websites that had opted out of scraping and then tried to circumvent controls after being blocked. With continuous scoring, deviations like that would surface in the trust band instead of waiting for a high-profile incident.

Rollout and the wider loop

For now, Cloudflare is running the system internally to build behavioral history for its Verified bot list and refine how observations become useful guidance. The company plans to publish the scores on Radar and to expand what it measures. It also positions the system alongside Bot Preference Sync, which helps site owners publish their preferences, and Bot Management, which decides how automated traffic is treated — a combination Cloudflare describes as an accountability loop for automated traffic. Eventually, all Cloudflare customers will be able to choose the trust threshold applied to Verified bot traffic.

Why it matters

Verified status has so far been a statement about identity, and identity alone is a weak guarantee once a crawler is inside the gates. Tying the privilege to sustained, measurable, explainable — and eventually public — behavior changes the incentives for every major operator of automated traffic. Site owners get evidence-based grounds for allowing bots and a dial for tightening the rules, while responsible operators get a way to demonstrate they deserve the access they rely on. As the Perplexity episode showed, even prominent players can flout scraping opt-outs, which makes verifiable ongoing accountability arguably more valuable than verification itself.