AI fetch timeout risk checker
The AI fetch timeout risk checker requests your page three times, once each with the ChatGPT-User, Claude-User and Perplexity-User user agents, and records the HTTP status, total time and HTML size. The median time sets the risk: low at 500 ms or under, moderate up to 1.2 seconds, high above that.
- Fetches
- 3, one per AI user agent
- Risk from
- the median time
- Cost
- free, no signup
Three fetches
One row per AI user agent, with status, time and HTML size
-
Not runChatGPT-User—
Why it mattersOpenAI's agent for fetching a page when a ChatGPT user asks about it or shares a link. The checker sends its user-agent string, so a server that treats it differently will show it here.
How to fixIf this fetch is slower than the others or returns 403, check firewall, CDN bot rules and rate limits for this user agent, and cache the page so it can be served without heavy server work.
-
Not runClaude-User—
Why it mattersAnthropic's agent for fetching a page when a Claude user asks for it. A blocked or slow response here means the page may not reach the answer.
How to fixConfirm robots.txt and bot-protection rules allow Claude-User if you want it to read the page, then reduce time to first byte with caching.
-
Not runPerplexity-User—
Why it mattersPerplexity's agent for fetching a page on a user's request. Like the others, it waits for the HTML to arrive before it can use any of it.
How to fixServe the main content in the initial HTML, keep the page weight down and avoid redirect chains so the response completes quickly.
Fetch runs
Status, time and HTML size for each user agent
Nothing to show yet. After a test, a table of the three fetches with HTTP status, time in milliseconds and HTML size in bytes appears here.
How the risk is rated
The checker fetches the page three times, up to 2 MB each, with a 12-second limit and up to three redirects. The risk band comes from the median of the three times, so one slow outlier does not decide the result.
| Median fetch time | Risk | Per-fetch row |
|---|---|---|
| 500 ms or under | Low | Pass |
| 501 to 1,200 ms | Moderate | Needs work |
| Above 1,200 ms | High | Fail (also on an error or HTTP 400 and above) |
The HTML arrives quickly. Timeout is unlikely to be the problem.
Usually fine, but redirects, edge challenges or heavy server work could tip it over.
A live fetch may give up before the page arrives.
The thresholds are MoxSEO's working cut-offs, not timeouts published by OpenAI, Anthropic or Perplexity. Treat the band as directional.
Why does fetch speed matter for AI answers?
When someone asks ChatGPT, Claude or Perplexity about a page, the assistant can send a user-triggered agent to fetch it live. That request has to finish while the user waits. Slow or unstable delivery makes live retrieval less reliable, especially when a page adds redirects, edge challenges, heavy server work or intermittent errors.
The vendors document these agents: OpenAI's crawlers and user agents and Perplexity's crawlers describe ChatGPT-User and Perplexity-User. Speed is only one condition. The agent must also be allowed in, which you can check with the AI Crawler Access Checker, and the content must be in the HTML rather than added by JavaScript.
A fast HTML response does not prove that important content is present without client-side JavaScript, and a slower result does not identify the cause by itself. It is a prompt for a reproducible technical investigation.
For the background, read why AI assistants drop slow sites.
How to use the result
-
Step 1
Test the canonical public URL
Use the final page people and crawlers should reach. Redirect chains add latency and make the first request look different from the destination.
-
Step 2
Compare the three rows
If one user agent is much slower, or returns 403 or 429 while the others return 200, your server or CDN is treating that agent differently.
-
Step 3
Separate server response from rendering
Check the HTML size. A very small page that loads fast may be an empty JavaScript shell, which most AI fetchers cannot read.
-
Step 4
Retest before changing infrastructure
Repeat the check at different times. Then review origin response, CDN caching, security rules and application behaviour.
-
Step 5
Confirm with your own logs
Look for requests from these user agents in your server logs, and for abandoned requests such as status 499 in nginx, which mean the client gave up before the response finished.
Directional, not a monitor
This checker does not measure Google ranking factors, guarantee AI citation or replace continuous monitoring. It runs from one server location, not from the real AI fetchers' IP addresses, so geography and IP-based rules can make real fetches faster or slower.
What the checker does
- Fetches the URL three times with real AI user-agent strings
- Follows up to three redirects, 12-second limit, up to 2 MB
- Records HTTP status, total time and HTML size
- Rates risk from the median time
What it does not do
- Fetch from OpenAI, Anthropic or Perplexity IP addresses
- Measure Core Web Vitals or uptime
- Run JavaScript or render the page
- Check robots.txt (use the AI Crawler Access Checker)
Need help investigating slow responses?
Review response timing, redirects and server constraints before deciding which performance work to prioritise.
AI fetch timeouts, answered plainly
Short, standalone answers. Email editorial@moxseo.com if any of them is wrong.
What does the AI Fetch Timeout Risk Checker measure?
It fetches a public URL three times from MoxSEO's server, once each with the ChatGPT-User, Claude-User and Perplexity-User user-agent strings. For each fetch it records the HTTP status, total time in milliseconds and HTML size, then rates timeout risk as low, moderate or high from the median time.
What counts as low, moderate and high risk?
Low risk is a median fetch time of 500 milliseconds or under. Moderate is 501 to 1,200 milliseconds. High is anything above 1,200 milliseconds. These are MoxSEO's working thresholds, not published limits from the AI companies.
Is this the real ChatGPT or Claude fetching my page?
No. The requests come from MoxSEO's server using the same user-agent strings. That shows how your server responds to those agents, including any user-agent blocking, but not how it treats their real IP addresses.
Why is one user agent blocked when the others are not?
Firewalls, CDN bot rules and security plugins often match on user-agent strings. If one row shows 403 or 429 while the others show 200, a rule is treating that agent differently. Decide whether you want it blocked and adjust the rule.
My page is fast but still not cited. Why?
Speed only removes one obstacle. The agent must also be allowed by robots.txt and bot protection, the content must be in the HTML rather than rendered by JavaScript, and the page must be structured so an answer can be quoted. Citation also depends on how the assistant chooses sources.
How can I confirm a real timeout?
Check your server logs for requests from these user agents. Abandoned requests, such as status 499 in nginx, mean the client closed the connection before the response finished.
Run the next check
All 24 free tools- AI Crawler Access CheckerA fast response is no use if robots.txt blocks the agent.
- AI Answer Extractability CheckerOnce the page arrives, can a model quote a clean answer?
- On-Page SEO AuditRedirect chain, status and indexability for the same URL.
- Common Crawl Visibility CheckerSee whether Common Crawl captured the domain recently.
Slow for AI, slow for everyone?
Let us look.
A free 30-minute call with Sakshi Kumari or Ashish Khan. Bring this result; they will connect access, latency and extractable answers into one AI search plan.
A call is not an engagement. We never guarantee rankings or AI citations.