Skip to main content
Plans: All plans · In the app: Sidebar → Actions → Opportunities
Quotable checks each monitored brand’s site about once a day. Every check is a fixed rule over what your site sent back, with no AI model involved, so an unchanged site gives the same result every time and each finding points to the response, header or robots.txt line that caused it. Each failing check becomes one item in the Fix how AI reads your site group on Opportunities, filed under the Technical type. Passing checks are only counted.

How the check runs

  • How often. About once a day per brand. Brands that have never been checked go first.
  • Which brands. Every monitored brand that has a domain. Paused brands and ended pitches are skipped, as they are for daily tracking.
  • What it requests. About a dozen requests to the brand’s own domain, and nowhere else:
    • robots.txt
    • up to three declared sitemaps, plus one sitemap inside a sitemap index
    • the homepage, fetched several ways: as a plain client, as an AI agent, a few times in a row for timing, and as Markdown
    • /llms.txt
    • one address that cannot exist, to see your “not found” response
  • From where. From our network, under the name QuotableBot.

Identify and allow QuotableBot

Most requests carry this user agent:
One homepage request uses an AI-agent user agent to see how your firewall treats AI agents. It ends with QuotableBot/1.0, so your logs can tell it apart. The site check reads your robots.txt and reports what it allows rather than following it, and it only requests your own domain. On other people’s sites, QuotableBot follows robots.txt and answers to User-agent: QuotableBot. See QuotableBot for how it behaves on any site.
If your firewall challenges unknown bots, allow QuotableBot by user agent. Otherwise the checks report the challenge page rather than your site.

What is checked

Impact is High when the problem stops engines from reaching or reading your site, or makes them read it badly. It is Medium when fixing it improves how they read you. Titles are shown as they appear on the board.

Server and access

Blocking training crawlers such as GPTBot, ClaudeBot, Google-Extended, Applebot-Extended and Meta-ExternalAgent is your choice. The check leaves it alone.

Indexing

Structured data

Files for agents

A check that lacks what it needs is skipped, not failed. For example, the heading checks are skipped when the homepage did not return HTML. Skipped checks do not count toward the total.

Read a result

Open a Technical item to see:
  • What this is based on: what the check found, such as the URL tested, the status it got or the rule that matched. The last line, Checked, is the day of the check.
  • The change: for checks that can say exactly, what is there Now and what to Change it to. For example, your robots.txt group with a suggested Content-Signal line added, your heading outline corrected, or a starter llms.txt.
  • How to do it: the steps to fix it.
With Technical as the only type filter and nothing failing, the list says “All N site checks passed.” Before the first check it says “No site check has run yet.”

Crawl fixes from Agent Analytics

The site check shows what a visit would find. When the brand has an active Agent Analytics source, the same group also lists what AI systems did hit on real visits over the last 30 days: Open the visits takes you to the matching visits in Agent Analytics. These items leave the board on their own once later visits stop showing the problem.

Fix and confirm

1

Work on it with Quotable Agent

Open the item and click Work with agent. Quotable Agent opens beside the page with the item loaded. It asks about your setup, such as your CDN or CMS, then writes the file, config or copy for your brand. You need edit access to the brand.
2

Let the agent check the live page

Quotable Agent reads pages on your brand’s domain live: status, redirects, noindex, canonical, title, meta description, Open Graph tags, headings, files like robots.txt and llms.txt, and signs of a bot challenge or JavaScript-only content. It can also ask for a page as Markdown. It checks before it advises and again before it treats a fix as live. If it still sees the old version, it asks whether the change was published, since caches can lag.
3

Mark it Done on the day it went live

Set the status to Done and set Done on to the day the change went live. Quotable Agent can also propose the change, which you approve on a card. The Impact page puts its marker on that day.
4

Wait for the next daily check

The item leaves the board once a later check passes. There is no button to run the full site check on demand.
If the brand uses Agent Analytics, the Crawlability tab can also re-check robots.txt (Check again) and test one page (Test page) right away. See the Agent Analytics dashboard.

Limits

  • The daily check looks at the site as a whole and at the homepage. It does not crawl every page.
  • Quotable Agent checks up to six pages per reply, on the brand’s domain and its subdomains only. Its checks share the brand’s limit of 30 diagnostics per hour with the Agent Analytics tests.

Next steps

Opportunities

Work through the items with your team.

Impact

See what moved after a fix went live.

Agent Analytics

See what AI systems hit on real visits.
Last modified on November 18, 2026