Allow or block AI crawlers? Weighing it up

The decision usually gets argued as a matter of principle and is a business question: if you earn from content, allowing costs you. If you win projects through content, blocking costs you — and more.

A row of narrow apertures, light streaming through three of them into the space beyond while the rest stay dark

In short

  • There are two kinds of request: advance collectors and runtime fetches. They can be controlled separately — and that is where the most usable solution lies.
  • If you sell projects, allow the runtime fetches in any case: they produce mentions in front of people mid-way through looking for a solution.
  • If you sell content or earn from ad impressions, there are good reasons to block the collectors.
  • robots.txt is a request, not a block. Enforcing it technically requires a rule at server level.

The two kinds of request

Advance collectors

Fetch content independently of any specific question — for training or for a search index of their own. Examples: GPTBot, PerplexityBot, ClaudeBot, Google-Extended, Applebot-Extended, CCBot.

Benefit to you: indirect — your content sits available in an index. Cost: your content gets used without anyone visiting your page.

Runtime fetches

Fetch a page at the moment someone has asked a matching question. Examples: OAI-SearchBot, ChatGPT-User, Perplexity-User, Claude-User.

Benefit to you: direct — mentions and links in specific answers come from these. Cost: practically none.

Worth knowing

Blocking "all AI bots" blanket blocks both kinds — and thereby loses exactly the part that benefits you. Being named as a source in an answer requires the page to be fetchable at runtime.

For a company selling projects that is the worst possible trade: you prevent a recommendation reaching someone mid-way through looking for a solution, in order to prevent a model learning from a publicly accessible article.

Three strategies

StrategyCollectorsRuntime fetchesSuits
Openallowallowservice businesses, B2B, consulting
Mixedblockallowpublishers, course providers, trade media
Closedblockblockpaid content, protected areas

For most small and mid-sized companies, "open" is the right choice. The mixed strategy is the sensible middle way for anyone whose content is itself the product.

The robots.txt

For the open strategy with explicit naming — not strictly necessary, but it makes the decision visible and documents it:

robots.txt – open
User-agent: GPTBot
Allow: /

User-agent: OAI-SearchBot
Allow: /

User-agent: ChatGPT-User
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Perplexity-User
Allow: /

User-agent: ClaudeBot
Allow: /

User-agent: Claude-User
Allow: /

User-agent: Google-Extended
Allow: /

User-agent: Applebot-Extended
Allow: /

User-agent: *
Allow: /

Sitemap: https://example.com/sitemap.xml

For the mixed strategy, the collectors get blocked and the runtime fetches allowed:

robots.txt – mixed
User-agent: GPTBot
Disallow: /

User-agent: ClaudeBot
Disallow: /

User-agent: PerplexityBot
Disallow: /

User-agent: Google-Extended
Disallow: /

User-agent: Applebot-Extended
Disallow: /

User-agent: CCBot
Disallow: /

User-agent: OAI-SearchBot
Allow: /

User-agent: ChatGPT-User
Allow: /

User-agent: Perplexity-User
Allow: /

User-agent: Claude-User
Allow: /

User-agent: *
Allow: /

Sitemap: https://example.com/sitemap.xml
Careful Google-Extended controls only the use of content for Google's AI products — not ordinary search indexing. Block Googlebot by accident and you disappear from search. Those two names are different things and get confused regularly.

What robots.txt does not do

Three limitations worth knowing before relying on the file:

  • It is a request. Reputable providers observe it, others do not. Enforcing it technically requires a block at server or firewall level based on the user agent.
  • It does not work retroactively. What has already been collected does not disappear because of it.
  • It does not protect against reproduction from other sources. If your content gets quoted elsewhere, your own file does not help.
From practice

Before making the decision, it is worth looking at the server logs: which of these identifiers appear at all, how often, and which pages do they fetch?

On small websites the answer is often sobering — few requests, few pages. The question is then practically secondary, and the actual task is a different one: getting found at all. A block against crawlers that are not coming solves no problem and merely prevents that changing later.

The legal frame

In the EU a text and data mining reservation exists: rights holders can expressly prohibit the use of their content for automated analysis, and that reservation has to be declared machine-readably. The robots.txt is the common form for that, sometimes supplemented by a note in the terms of use.

It does not apply directly in Switzerland; anyone also addressing EU users should nonetheless declare the reservation if they want it. That is a point for a legal professional to assess — this piece does not replace legal advice.

In closing

The question is not one of principle but hangs on the business model. If you sell content, block the collectors. If you sell projects, allow both — and gain mentions in front of people currently looking for a solution.

In either case: consider the runtime fetches separately. Blocking them along with the rest is the most common and most expensive mistake in this decision.

Common questions

Should you block AI crawlers?

It depends on the business model. Anyone earning from content itself — publishers, course providers, trade media — has good reasons to block the advance collectors. Anyone selling projects or services loses more by blocking than they gain.

What is the difference between collectors and runtime fetches?

Collectors such as GPTBot, ClaudeBot or PerplexityBot fetch content independently of any specific question, for training or an index of their own. Runtime fetches such as ChatGPT-User, Perplexity-User or OAI-SearchBot fetch a page exactly when someone has asked a matching question — and mentions and links come from those.

What is Google-Extended?

An identifier controlling the use of content for Google's AI products — not ordinary search indexing. It is a different thing from Googlebot; block that one by accident and you disappear from search.

Does every crawler observe robots.txt?

No. It is a request, not a technical block. Reputable providers observe it, others do not. Enforcing a block requires a rule at server or firewall level based on the user agent — and even then it does not work retroactively on content already collected.

How do you check which crawlers are even coming?

Through the server logs, filtered by the known identifiers. That shows which crawlers fetch which pages, how often. On small websites the result is often modest — the blocking question is then secondary to the task of getting found at all.

Marketing that sets itself up

The Studio Engine beta is live. Claim your spot and help shape it from the start.

Join the beta →
← Back to overview