Verify every AI bot that says it is GPTBot
AI bot verification checks that a request claiming to be GPTBot, ClaudeBot, PerplexityBot or another AI crawler really came from that company. Proofsource recognizes AI crawlers and on-demand fetchers by name, proves each request against the operator's published IP ranges and reverse DNS, and tags every bot by intent: training, search or user fetch. Site owners and SEO teams get crawler numbers they can trust and robots decisions based on what each bot actually does.
| User agent | IP | IP range | Reverse DNS | Verified |
|---|---|---|---|---|
| GPTBot | 20.171.207.x | In OpenAI range | Not required | Yes |
| GPTBot | 185.220.101.x | Outside range | No match | No |
| ClaudeBot | 160.79.104.x | In Anthropic range | Not required | Yes |
| PerplexityBot | 45.152.66.x | Outside range | No match | No |
| Googlebot | 66.249.66.x | In Google range | crawl-66-249-66-x.googlebot.com | Yes |
| OAI-SearchBot | 52.230.152.x | In OpenAI range | Not required | Yes |
AI bot verification, in short
The facts, ready to quote.
- Recognizes
- AI crawlers and on-demand fetchers by user agent name, such as GPTBot, ChatGPT-User, ClaudeBot and PerplexityBot
- Verifies by
- The operator's published IP ranges and reverse DNS
- Result per request
- Verified or not verified, never assumed from the user agent alone
- Intent tags
- Training crawler, search crawler or on-demand user fetch
- Why it matters
- Spoofed bots inflate crawler counts and can hide real blocks
- Used in
- Crawler traffic reports and AI crawler access checks
AI bot verification, explained
Every web request carries a user agent string, a short label the client sends to say what it is. AI companies publish the names their crawlers use, such as GPTBot and OAI-SearchBot for OpenAI, ClaudeBot for Anthropic and PerplexityBot for Perplexity. The catch is that anyone can type any user agent. Scrapers, SEO tools and abusive bots routinely pretend to be GPTBot or Googlebot because many sites let those names through. A crawler report that trusts the user agent alone can count thousands of visits that never came from an AI company.
AI bot verification closes that gap with two checks the operators themselves document. The first is the IP range: OpenAI, Perplexity and others publish the address ranges their crawlers send from, so a request from outside those ranges is not the real bot. The second is reverse DNS: the request's IP address is looked up to a host name, the host name must belong to the operator's domain, and a forward lookup of that host name must return the same IP. A request that passes is the real bot. A request that fails is a spoofed visitor using a borrowed name.
Verification also needs intent. One company can run several bots with different jobs. GPTBot collects content that may be used for model training, OAI-SearchBot fetches pages for ChatGPT search results, and ChatGPT-User fetches a page when a person's conversation needs it. Proofsource recognizes each AI bot by name, verifies it by IP range and reverse DNS, and tags it as a training crawler, a search crawler or an on-demand user fetch, so you can allow search and user fetches while making a deliberate choice about training.
How it works
Tell real GPTBot, ClaudeBot and PerplexityBot visits from spoofed ones, and know whether each bot trains, searches or fetches for a user.
Recognize the bot by name
Each request's user agent is matched against known AI crawler and on-demand fetcher names, so GPTBot and ChatGPT-User are told apart, not lumped together as OpenAI.
Check the published IP ranges
The request's address is compared with the IP ranges the bot's operator publishes for that crawler.
Confirm with reverse DNS
Where an operator documents it, the IP is looked up to a host name on the operator's domain and checked back to the same IP.
Tag the intent
Every verified bot is tagged as a training crawler, a search crawler or an on-demand user fetch.
Report verified traffic
Crawler reports and access checks use the verified result, and requests that fail verification are shown apart instead of counted as AI visits.
See it in the product
Screens from the Proofsource app, drawn with sample data for an example brand.
| Bot | Operator | Intent | Verified visits |
|---|---|---|---|
| OAI-SearchBot | OpenAI | Search crawler | 4,120 |
| GPTBot | OpenAI | Training crawler | 3,880 |
| ChatGPT-User | OpenAI | On-demand user fetch | 1,310 |
| PerplexityBot | Perplexity | Search crawler | 2,640 |
| Perplexity-User | Perplexity | On-demand user fetch | 470 |
| ClaudeBot | Anthropic | Training crawler | 2,210 |
Every capability, live today
3 shipped capabilities behind AI bot verification, each in the product now.
AI bot recognizer and intent tags
Recognizes AI crawlers and on-demand fetchers by their user agent names, then tags each as a training crawler, a search crawler or an on-demand user fetch, the distinction your robots.txt decisions depend on.
AI bot verifier
Proves a request really came from the bot it names by checking the operator's published IP ranges and reverse DNS, so spoofed visits are kept out of your numbers.
ProofBot explained
A page describing what the ProofBot user agent reads and how to allow it on your site.
Who it is for
Block the fakes, keep the real bots
See which requests borrow an AI bot's name from addresses its operator does not use, and rate limit those without touching the real crawlers.
Report AI crawler numbers you can defend
Show leadership verified AI crawler visits instead of raw user agent counts that include scrapers.
Decide on training with the facts
Separate training crawlers from search crawlers and user fetches before choosing what to allow in robots.txt.
Audit a client's bot traffic
Explain to a client why their logs show more GPTBot than OpenAI actually sends, and which share of AI traffic is real.
Why it matters
Numbers you can trust
Only verified requests count as AI crawler visits, so a spike means real bots, not someone borrowing a name.
Robots choices by intent
Training, search and user fetch are separate decisions. Intent tags make each one visible.
Safer allowlists
Allow real AI crawlers by verified identity instead of by a user agent string anyone can send.
Clear unknowns
When an operator publishes no range or DNS rule, the result says not determined rather than guessing.
Google AI OverviewsGoogle AI Mode
AI bot verification: Proofsource and other tools
Several AI visibility tools report AI crawler visits. This table lists what each one's public pages describe about recognizing and verifying those bots.
| Capability | Proofsource | Profound | Peec AI | Otterly.AI | Scrunch | Ahrefs | Semrush |
|---|---|---|---|---|---|---|---|
| AI crawlers recognized by name | YesAI crawlers and on-demand fetchers recognized by user agent, each counted separately | YesWhich AI bots access content, how often, real-time bot monitoringSource | YesBot visit insights via CDN and log providersSource | YesWhich AI crawlers and agents visit your site, from server logsSource | YesReal-time feed of AI bot crawls and traffic trendsSource | YesBot Analytics shows real-time AI crawler visits and top pagesSource | Not listed |
| Genuine bots told from spoofed ones | YesPublished IP ranges and reverse DNS, verified or not verified per request | YesSeparates genuine AI bots from spoofed crawlersSource | Not listed | Not listed | Not listed | Not listed | Not listed |
| Intent per bot | YesTraining crawler, search crawler or on-demand user fetch | Not listed | Not listed | Not listed | Not listed | Not listed | Not listed |
| Check that AI bots can reach your site | YesVerified bot results feed the AI crawler access checks | Not listed | RelatedSitemap, llms.txt and robots.txt crawlability historySource | YesChecks AI bot access; works on competitor sites tooSource | YesMonitors crawl health for AI botsSource | Not listed | YesFlags blocked AI bots including ChatGPT-User, OAI-SearchBot, PerplexityBot, Claude-User, Claude-SearchBotSource |
Where each of them shines
Profound's Agent Analytics combines real-time bot monitoring with crawler verification and log forwarding from many CDNs. Compare in full
Peec AI brings bot visits in from CDN and log providers including Cloudflare, CloudFront and Akamai. Compare in full
Otterly.AI offers several easy log connections for agent analytics, including upload, webhook, Cloudflare and a WordPress plugin. Compare in full
Scrunch pairs a real-time bot feed with crawl health monitoring and crawl error detection. Compare in full
Ahrefs offers Bot Analytics with top pages, free in beta, next to its web analytics. Compare in full
Semrush Site Audit flags which AI bots a site blocks, including search and user-fetch bots, next to its AI Search Health checks. Compare in full
Questions about AI bot verification
How do I know if a GPTBot visit is really from OpenAI?
Check the request's IP address against the IP ranges OpenAI publishes for GPTBot. A request from outside those ranges is not the real GPTBot, whatever its user agent says. Proofsource runs this check on every request that claims to be an AI bot.
What is reverse DNS verification?
Reverse DNS looks up the host name behind an IP address. For a bot to pass, the host name must belong to the operator's domain and a forward lookup of that name must return the same IP. It is the method search engines such as Google document for verifying their crawlers.
Why would anyone fake an AI crawler's user agent?
Many sites and firewalls let known crawlers through, so some automated clients borrow names like GPTBot or Googlebot to get through. The result is inflated crawler counts and allowlists that let unwanted traffic in.
What is the difference between a training crawler, a search crawler and a user fetch?
A training crawler, such as GPTBot or ClaudeBot, collects content that may be used to train models. A search crawler, such as OAI-SearchBot, Claude-SearchBot or PerplexityBot, fetches pages for answers with live sources. A user fetch, such as ChatGPT-User, happens when a person's conversation asks the assistant to read a page.
Does blocking GPTBot remove my site from ChatGPT answers?
No. GPTBot is OpenAI's training crawler. ChatGPT search results use OAI-SearchBot, and ChatGPT-User fetches pages on request. Blocking GPTBot is a training decision; blocking OAI-SearchBot affects whether you appear in ChatGPT search.
Which AI bots does Proofsource recognize?
Proofsource recognizes the AI crawlers and on-demand fetchers of the major answer engines by name, including GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot and PerplexityBot, and tags each by intent.
What happens when an operator does not publish IP ranges?
The request is marked not determined rather than verified or failed. Proofsource never guesses a verification result.
How is this different from Profound's crawler verification?
Profound's Agent Analytics also separates genuine AI bots from spoofed crawlers, alongside real-time monitoring and broad log forwarding, which is a real strength. Proofsource adds an intent tag per bot and uses the verified result in its crawler access checks as well as its traffic reports.
Keep exploring
Try it on your own brand.
Proofsource is live: scan your brand and see AI bot verification at work with your own answers, competitors and sources.
Google AI OverviewsGoogle AI Mode