Lectern's access checks

On a customer's own site, Lectern requests a page under each assistant's product token, so the site's own rules for that assistant apply, and compares the answer with a browser's. Each of these user agents names Lectern and links to this page:

• ChatGPT-UserMozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; ChatGPT-User/1.0; LecternAccessCheck/1.0; +https://trylectern.com/bot
• Claude-UserMozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; Claude-User/1.0; LecternAccessCheck/1.0; +https://trylectern.com/bot)
• Perplexity-UserMozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; Perplexity-User/1.0; LecternAccessCheck/1.0; +https://trylectern.com/bot)

These requests come from Lectern's own servers. They are not sent by OpenAI, Anthropic or Perplexity, and they are not those assistants. A firewall that verifies bots by address or signature will rightly treat them as unverified.

Only sites whose ownership a Lectern customer has verified: the site's own property in Google Search Console, connected to Lectern. Google grants that access only after the owner proves control of the domain.

In these access checks, a publisher or other third-party site Lectern tracks is never asked under an AI assistant's name or a browser's. It gets robots.txt and one page request, both as Lectern's crawler:

A customer site that is not verified, and any site entered in our free access check, gets robots.txt, Lectern's own read and one browser request. Nothing is sent to it under an AI assistant's name.

Lectern's own reads of robots.txt and of each page, in the daily check, the free check and a site audit, use this user agent:

On a customer's site and in the free check, the browser request each comparison is made against uses an ordinary desktop browser's:

• A customer's site: once a day. robots.txt, then the homepage and up to two of its most visited pages. On a verified site each page gets one request as Lectern, one as a browser, and one as each of the three assistants, one second apart: at most 16 requests a day.
• A site audit, when a customer starts one: the audited page, robots.txt and about 35 small requests for well-known files (llms.txt, the sitemap, an OpenAPI file and the like), all as LecternAccessCheck. On the customer's own verified site it then asks the audited page once as a browser and once as each assistant.
• A publisher where a customer has or is considering a placement: at most once a day, robots.txt and one page request, both as LecternCrawler, which keeps to the rules robots.txt gives it.
• The free access check: when someone asks, robots.txt, the homepage as Lectern and once as a browser. A complete answer is reused for five minutes; how often one person can ask is limited.

They never sign in, submit forms, solve challenges or retry around a refusal, and requests named after an assistant, and the browser request they are compared with, never follow a redirect to another site. When a site refuses Lectern's own robots.txt or page read, nothing more is sent to that site in that check. When it refuses an assistant's or the browser's request, nothing more is sent to that page, and the refused user agent is not sent again in the check. An audit stops asking once the site rate-limits or challenges it.

Email us at hello@trylectern.com with the domain, and we will stop sending these requests to it. You can also refuse the user agents above in your firewall or CDN; Lectern does not retry after a refusal.