Should You Block GPTBot ClaudeBot PerplexityBot on WordPress?
Most WordPress owners should not block GPTBot ClaudeBot PerplexityBot across the board, because a blanket block quietly removes the site from AI answers. Each crawler is a separate opt in, so a single robots.txt rule never covers all 3. The decision comes down to whether you care more about keeping content out of model training or staying citable in AI search.
Quick answer: GPTBot, ClaudeBot and PerplexityBot are 3 distinct user agents, and each needs its own robots.txt rule. Blocking a training agent keeps your writing out of model training data. Blocking a search or answer agent deletes you from AI citations, which is the outcome most businesses do not actually want. PerplexityBot is search oriented, so allowing it is the usual recommendation for brands that want AI visibility. robots.txt is a request rather than a wall, so server or firewall rules are the only hard control, and access logs are the only place to confirm what really changed.
What GPTBot, ClaudeBot and PerplexityBot Actually Do
AI crawlers fetch pages to power AI search and to build training data. The 3 that matter most on WordPress are GPTBot from OpenAI, ClaudeBot from Anthropic and PerplexityBot from Perplexity. Every 1 of them announces itself with a distinct user agent string, so your access logs tell you exactly which bot arrived, which page it asked for and when.
These are independent opt ins. A rule that stops GPTBot does nothing to Google-Extended, ClaudeBot or PerplexityBot. If the goal is to keep every AI crawler off the site, you need a separate block for every user agent you want to stop, and you need to keep that list current as vendors add new agents.
Why Site Owners Block AI Crawlers
The usual motive is training. If your articles, data and images feed a model without payment or credit, a robots.txt block on the training agent is the only lever most site owners have without touching server config.
There is a second motive that gets confused with the first. Some vendors run separate user bots that fetch a live page so a chat interface can summarize it. Blocking those stops AI tools from quoting or linking your live pages inside a conversation. That is a visibility decision, not a copyright decision, and the 2 rarely deserve the same answer. The middle path, blocking scrapers without losing AI search visibility, is where most publishers end up.
The Real Risk Is Blocking the Wrong Bot
Here is the uncomfortable part. When a crawler cannot reach a page, the AI engine cannot cite it. The answer still gets written. Your site is simply absent from it, and no report will tell you about the traffic you never received.
A blocked training agent means your content stays out of model training. A blocked search or answer agent means you lose the citation, the brand mention and any referral click that came with it. You can block training access and still allow answer access, but only if you know which user agent does which job. That clarity comes from logs, not from guessing at names.
How to Block GPTBot, ClaudeBot and PerplexityBot in WordPress
Start by deciding what you are protecting. If the worry is training, target the training agents. If the worry is live summaries, target the user agents. Then write the rules:
- Open robots.txt through your SEO plugin, your hosting panel or a code editor.
- Write a separate user agent block for GPTBot, ClaudeBot, PerplexityBot and any other crawler you want stopped.
- Add llms.txt as a companion if you like. robots.txt controls access; llms.txt points AI engines at your best material.
- Check the access logs a week later and confirm the bots you blocked stopped arriving.
Remember that robots.txt is a polite request. Blocking at the server or firewall level is the only control that holds for a badly behaved crawler. A fuller treatment of allow lists, rate limits, and log review covers the controls that sit behind that file. Redirects deserve a test too. These 3 agents handle HTTP redirects differently, with tolerance for roughly 3 to 5 hops observed in practice, while the specification allows 30 and Googlebot allows 10. A redirect chain that Google walks without complaint can lose an AI crawler entirely.
A Decision Framework for WordPress Owners
The choice reduces to what you value more: keeping content out of training, or keeping the site inside AI answers.
| Your priority | Recommended move | What it costs you |
|---|---|---|
| Keep content out of model training | Block the training agents such as GPTBot and ClaudeBot | Nothing in citations, provided the answer agents stay allowed |
| Stay visible in AI answers | Allow the search and user agents, including PerplexityBot | Your pages can be summarized and linked inside chat interfaces |
| Hard control regardless of robots.txt | Block at the server or firewall level | A wrong user agent in the rule set wipes you out of AI answers |
Most businesses land in the middle. They stop the training crawlers to protect original work, then leave the answer crawlers alone so pages can still be cited. Because the split between training and answering is not obvious from the agent names, match each agent to an outcome you can see in your own logs before you commit.
What Your Server Logs Tell You
Logs are the ground truth. They show which pages GPTBot, ClaudeBot and PerplexityBot requested, how often and with what status code. Without that record you are deciding blind, and a robots.txt edit becomes an act of faith.
If a bot hits the homepage 40 times a day and never produces a citation or a referral, blocking it costs you nothing. If a bot crawls your best guides and your brand keeps appearing in AI answers, blocking it removes that visibility overnight. Only data separates those 2 cases, which is why it pays to track AI crawler visits with the citAEOtion plugin and watch the pattern per bot and per page before and after every rule change. Book a live demo to see that pattern on your own site before you write the first rule, and read how to block AI crawlers for the three enforcement layers.
No single answer fits every site. A content heavy blog that earns citations usually wants the answer crawlers allowed. A client portal with no reason to appear in a chat window can be locked down at the server. Let the logs, not the headlines, decide which of those describes your site.
Frequently Asked Questions
Should I block ClaudeBot?
Blocking ClaudeBot keeps your pages out of Anthropic model training, which makes sense if protecting original writing is the priority. If you want to be cited in Claude answers, leave the matching search and user agents allowed. Check the logs first and see whether ClaudeBot visits your site at all.
How do I stop an AI bot completely?
Begin with a separate robots.txt rule for each crawler, since a block on GPTBot does nothing to ClaudeBot or PerplexityBot. For stronger control, block at the server or firewall level, because robots.txt is a request rather than a guarantee. Confirm the result in your access logs a few days later.
Does blocking GPTBot remove me from ChatGPT answers?
It can, depending on which agent you stop. The training agent and the live fetching agent serve different jobs, so a rule aimed at training can still leave live citations intact. Match each user agent string in your logs to the behavior you see before you assume the block is safe.
Should I block AI crawlers at all?
Most businesses should not, because a broad block removes the site from AI search answers entirely. The safer path is to separate training crawlers from answer crawlers, block only the agents whose activity hurts you and verify the effect in your logs. A total block is a visibility decision dressed up as a protection strategy.