# Known Limitations (/docs/known-limitations)


Some sites carry bot walls so strong that even the paid browser agent cannot clear them.
**Check this page before you spend credits** on a browser\_agent run against a site known
to be hard-walled. These blocks are not bugs — they are deliberate business decisions by
the site to refuse automated access entirely, and Bowmark respects that.

## Sites with complete bot-wall blocks

The following sites are **fully blocked** even with the paid `browser_agent`, meaning
every run will report `status: "failed"` with an error starting `blocked: `:

### Retail & eCommerce

| Site                         | Wall     | Why it's hard                                                                    |
| ---------------------------- | -------- | -------------------------------------------------------------------------------- |
| **fnac.com** (French retail) | DataDome | Block includes headers and JavaScript challenges; persists across proxy rotation |
| **amazon.fr**                | AWS WAF  | One of the strictest retail walls; blocks even residential proxies               |

### How to tell if a site is blocked

When you hit a hard bot wall, `browser_agent.status()` returns:

```json
{
  "status": "failed",
  "error": "blocked: <site> is refusing the browser"
}
```

**This is not a temporary error.** Retrying will not help — the wall refuses all automated
access, and no proxy, model, or backend option can bypass it.

## What to do instead

* **For retail & ecommerce sites:** Look for a typed provider in the library. Fnac and Amazon have
  many providers (`amazon_fr`, `fnac`, etc.) that reach the data through faster, cheaper routes
  than a browser agent would.
* **For a site with no provider:** Check whether the site publishes an API, RSS feed, or other
  machine-readable surface. See [The implementation ladder § Rung 9](../docs/how-it-works)
  for ways to find non-obvious data sources.
* **If none of those work,** the site's policy is to refuse automation, and that boundary is
  correct to respect. Contact the site directly if you need data access.

## Sites where the browser agent is slow or expensive

Some sites do *not* hard-block, but hit a wall or challenge that costs extra time and money:

\| Site | Issue | Typical cost | Alternative |
\|---|---|---|
\| **LinkedIn** | Heavy bot challenges; CAPTCHA on many queries | $1.50–$3.00 per run | API via credential |

  The walls change, sometimes weekly. If you hit a wall on a site not listed here,
  reach out to [support@bowmark.ai](mailto:support@bowmark.ai) or check back — it may be documented by then.

## How this list is maintained

We test high-traffic sites regularly. If a site changes its wall policy, or if you discover
a new wall we have not documented, email [support@bowmark.ai](mailto:support@bowmark.ai) —
each report helps us keep this list current.