Troubleshooting

The usual problems and how to resolve them, grouped by where they show up.

Crawling and page discovery#

The crawl finds very few pages#

  1. Check the Sitemap card on the Overview. If it says Not found, publish a sitemap.xml: it is the most frequent cause and the highest-impact fix.
  2. Review the site's max pages. If you set it low for a trial, raise it.
  3. Confirm you have not exhausted your plan's pages per site limit.
  4. Check that robots.txt does not block the paths you care about.

The crawl fails#

The Crawl failed message is usually down to a site that does not respond, an invalid SSL certificate or a firewall blocking the crawler. Review the configuration warnings on the Overview, which usually name the specific cause.

Warning

If your site answers with a 302 and sets cookies before serving content, AgentFi's crawler handles it automatically, but external agents (GPTBot, ClaudeBot, PerplexityBot) will receive the home page instead of the page they asked for. It is a real problem, and it has to be fixed on the server: bots must receive the requested page without going through a cookie redirect.

SSL certificate warning#

If the sitemap was detected using relaxed TLS, AI bots with strict certificate validation may not be able to reach you. Review your domain's certificate chain: the intermediate certificate is almost always the missing piece.

Optimization#

Pages stuck on Optimize Failed#

  1. Use Retry all from Pages: many failures are transient.
  2. If it persists on specific pages, look at their content: nearly empty pages, or pages loaded entirely through JavaScript, do not give the model enough material.
  3. Check that you still have optimizations left in the monthly package.

I am not happy with the optimized content#

  • Use Update content on that row and describe the specific adjustment.
  • For a change that affects the whole site, edit the custom optimization rules in the Optimize All dialog: they are saved and applied from then on.
  • If the original changed, use Re-crawl source before optimizing again.

The optimization package ran out#

Manual optimization pauses until the next period. Auto-optimization keeps going, but starts spending tokens. The ways out are to wait for the quota reset, upgrade the plan, or narrow the scope by excluding pages that add nothing.

Integration#

Domain verification does not pass#

  1. Confirm the record is a TXT, not a CNAME or an A record.
  2. Confirm the name is @ or your bare domain.
  3. Check it from outside: dig +short TXT your-domain.com.
  4. Wait: propagation can take up to 24 hours, though it is usually a matter of minutes.

Cloudflare zone not found#

The token belongs to an account that does not manage that domain. Three ways out:

  • add a token from the Cloudflare account that does manage the domain;
  • add the domain to the Cloudflare account the token belongs to;
  • grant that token Zone Read permission on the domain.

A permission is missing on the Cloudflare token#

The dashboard names the missing permission. Open the token's settings in Cloudflare, add that permission and add the token again in AgentFi: a saved token does not pick up new permissions on its own.

Route conflict#

Danger

A Worker route already exists for the same pattern, assigned to another worker. Deploying will overwrite it. Check in your Cloudflare dashboard that the worker is not critical before ticking the confirmation box.

I installed Nginx or Apache and nothing happens#

Almost always because the Cloudflare Worker is still live and intercepts everything at the edge, so your origin configuration never sees the traffic. Follow the migration order described in Integration: install and reload the origin, confirm it answers, Disconnect Cloudflare, and only then Verify & Activate.

Proxy verification fails#

  1. Confirm the Worker is deployed and that the route is live on your domain.
  2. Use the panel's Test /verify endpoint button.
  3. Check the exact URL the panel says it checked: a base-path mistake is common.

The integration shows a warning#

Automatic monitoring runs every 10 minutes. Review the time of the last check and the error message. A one-off failure can be a momentary outage; a persistent one means the Worker was deleted, the route changed, or the origin is not responding.

Discovery files#

llms.txt is not generated#

If the mode is Managed by AgentFi but there is no optimized page, the file cannot be built. Optimize at least a few pages and try again.

The file does not show up on my domain#

  1. Check that the integration is active: managed files are served by the proxy.
  2. Check that the file's mode is Managed by AgentFi and not Use site's original.
  3. Click Re-detect and then open the URL directly in the browser.

Visibility#

Every query comes back Not mentioned#

  1. Review optimization coverage: if it is low, the models have little content of yours to cite.
  2. Confirm the integration is active and that Bot traffic shows 200-status visits.
  3. Give the crawlers time: optimized content takes a while to enter the models' indexes.
  4. Read the search queries used by engine in a query's detail: they will tell you which terms the model is searching with, and what you are not covering.

Results change between runs#

AI engines are not deterministic. The same query can return different results in two consecutive runs. Look at the trend across several runs, not at one isolated test.

Lots of Engine error results#

It means the provider did not return a usable answer: an outage, a timeout or a content filter. It does not penalize your percentage. If it repeats en masse on one engine for several days, it is a problem on the provider's side.

The Claude column is empty#

Claude AI search is exclusive to the Business plan. Business

Usage and billing#

Tokens run out much sooner than expected#

  1. Open Change history and filter by Scheduled: you will see the usage of each run.
  2. Review the auto-test cadence. Daily × 40 queries × 4 engines is around 4,800 tokens a month.
  3. Check whether auto-optimization drained the package and moved on to spending tokens.
  4. Set a monthly token cap on auto-optimization.

Extra usage shows up in red#

You have gone past the tokens included and the additional spending is billed at your plan's per-token rate. The sidebar shows the extra tokens and the amount. To stop it, pause the schedules or set a monthly cap.

Access#

I see No site access#

Sites exist in the organization, but your user has permission on none of them. Ask an owner or an admin to grant it under Team.

I cannot see the Admin section#

That is expected: it only appears for users with an admin role.

I cannot change plan#

If your plan is marked Admin assigned, it is managed by your organization's admins.