Skip to main content
An engine can only cite a page it can read. Site readability in every Snapshot report lists what stops AI engines reading your site. It sits apart from the Heralded Score and has no score of its own. Website audits and the Command Center’s Website section score your whole site with the Website score, and these checks feed their findings.

Which pages it checks

Heralded’s direct page reads support public HTTP and HTTPS addresses on ports 80 and 443. Private networks and other ports are blocked. The checklist covers every page of yours the crawl read. Legal pages and files count toward the page checks, though they get no editorial checks. The readiness checks in the Snapshot report and Website count the same pages from the same crawl. The subscription inventory reads your site’s own default pages, plus your topic markets’ local languages where the site has them. Same-language regional copies and declared language alternates count as one page. It starts from the pages your sitemap and a search index list, then reads the pages those pages link to, until none are left or it reaches the page limit. A markdown (.md) copy of a page and a route template a sitemap lists, such as /blog/[slug], are not pages and are not read. Heralded lists your site’s pages at most once every seven days, and every read inside that window starts from the same list. A Europe topic adds German, French, Italian, Spanish, Dutch, Polish, Romanian, Swedish and Danish where your site already has them. Heralded uses detected language roots and declared language alternates, matching by language rather than country. It reads listed pages and their links, without guessing translation addresses. An English-only site adds no pages. A page engines cited in this run’s answers passes, since the citation shows they read it. A citation from an earlier run doesn’t count: a page cited before that is blocked now fails. A robots.txt exclusion on a URL with a query string, such as ?add-to-cart=123, does not count as a blocked content page or generate an unblock action. Exclusions on paths without a query string still count. Heralded finds repeated content such as posts, news, events and jobs from WordPress, feeds, sitemap files, URL families and structured data. It reuses results from usable sitemaps and WordPress listings for seven days. It reads category and tag sitemaps again when finding generated archives. Translated counterparts share a collection category. Quick Audits read up to three items from each collection. Weekly reports read up to 100 items from each collection. Base and Full audits use their page limit. A collection’s sample starts with items linked from pages Heralded reads, then fills any remaining places from the sitemap’s most recently modified items. Addresses break ties when modification dates match or are absent. One-time audits read the homepage and cited pages first, then prefer the address you entered and its redirect host over other subdomains. Generated archives count separately from that limit. Heralded reads up to three distinct samples of each taxonomy, such as tags or authors, within the read’s selected languages. Redirects to the same archive count as one sample. It follows a collection’s numbered listing pages in order while its sample has room. A listing page Heralded does not read adds no links or click depths. Date addresses count as generated archives on WordPress. On other sites, Heralded reads them as ordinary pages. Your first site inventory has its own reading allowance. Later reports use the monthly allowance to refresh pages. If an inventory stops before it finishes, daily updates continue the same weekly report’s inventory. Pages already read stay counted and are skipped. Coverage includes every pass, so its pages read and pages found never decrease during a continuation. If a week’s read of your pages does not finish, the Audit keeps the last complete read. After the first inventory, weekly reports reuse your page content until your site signals a change or the content is 28 days old. Cited pages and key pages go first, then each category of pages in turn. Heralded checks robots.txt on every run. Settled actions still get their own live checks.

Website page counts

The Website score’s coverage counts pages Heralded read that hold an issue or an improvement, out of all pages read. Information findings do not count. A page listed in your sitemap but not read is in neither number. On Website Pages, With findings counts matching pages before pagination. If any matching page has no measurement, it shows a dash. Cited in AI answers counts matching pages cited at least once, and Pages AI can reach counts those whose access checks pass. Filters apply to all three figures. If a read stops before it finishes, Heralded leaves the sitemap-link check not measured. It takes no points off your Website score and marks no pages for missing links. The check needs a complete read to judge whether your sitemap lists pages with no incoming links. AI can read counts pages that load and pass their page access checks. Where a page has no access check, it follows the server and browser text comparison. A page that needs JavaScript or answers with an error does not count. The pages-with-findings trend counts each report’s saved site map with the same definition. A report with no saved map keeps its recorded count, with a break before it. The Website score trend also breaks when its findings rules change.

Website scope

For your subscription, Heralded judges hosts and path sections within your brand’s registrable domain and its subdomains. A section is a folder up to two levels deep, such as /docs or /docs/api. The domain your site redirects to also belongs to your brand. Each has a state:
  • Read: pages are read and scored as your website.
  • Counted only: pages count in your inventory without being read.
  • Excluded: pages stay out of your inventory.
You can change a state in Brand settings under Hosts and sections. Each host and section there shows how many of your pages it holds, and the box gives your pages on the site from the weekly report Website shows. The pages used to judge content coverage belong to hosts and sections in state Read, in the languages your inventory reads, and in the Site, Content or Support groups. Large collections contribute their sample. Listed pages, other languages and the rest of a sampled collection do not make that coverage unread. Robots exclusions, a root page’s noindex directive and login restrictions exclude a host or section. A folder that refuses access does not exclude its public pages. Heralded checks the listed pages before treating the section as requiring sign-in. Language parameters such as ?lang=ro stay on listed addresses, so those pages are read in their language. Your sitemap, links from your main site’s navigation or footer, and a shared page template identify your own site. An unlinked host carrying another product’s public pages can be listed instead. Host and path names break ties; conflicting evidence gets a model judgement. Heralded checks a limited number of sections on each run. Until a section is checked, its pages follow the section it sits in, so a new folder on your main site is read with the rest of it. A folder named for community or app pages, such as /questions, and a host not yet checked are listed until checked. Heralded reuses automatic scope verdicts for up to seven days and checks robots rules on the URLs each report discovers. Nested sections receive their own root checks, so a public parent does not admit a login or noindex subsection. Sections waiting for their root checks are listed and retried during later discovery. A URL stays out of inventory until its robots check completes. Links can add listed URLs to your inventory without reading them. Your host and section states include the signals, the decision’s time, whether rules, a model or a member decided, and each one’s pages. pages_in_site counts your current pages in hosts and sections that are Read or Counted only. Read them through GET /v1/brands/{brand_id}/website/scope or the get_site_scope assistant tool. These are current settings, independent of a report’s date. With permission to change brand settings, you can override a host or path section through PUT /api/domains/{tracked_domain_id}/website/scope, supplying host, path_prefix and state. A path prefix defaults to / for the whole host and matches complete path components. Member overrides survive later discovery and take precedence over automatic decisions beneath them. A more specific member override takes precedence over a broader one. Each brand keeps its own decisions, even when another brand tracks the same domain.

Page types

Heralded classifies your pages from their address, then from their content when it reads them. The types are Home, About, Contact, Careers, Landing, Blog index, Blog post, Product, Product page, Collection, Solution, Pricing, Customer story, Press, Documentation, Resource, Comparison, Legal, Utility and Other. A blog listing is Blog index; an individual article is Blog post. On a shop, a page selling one item is a Product page and a listing of items for sale is a Collection. Product covers pages that market a product or its features. Utility covers pages such as search, login and thank-you pages. You can correct a page’s type in its detail pane if your role lets you dispatch changes. When classification rules change, each report refreshes up to 200 page types and leaves the rest for later reports. Existing page facts remain available.

What counts as a blocker

Each blocker comes with one line on how to fix it. The JavaScript blocker compares the same server and browser text as the Website finding. Pages that matter and share the same blocker raise one action, with the fix and every affected page in its checklist. Actions says which pages matter. A page Heralded couldn’t fetch is counted as not read, never as a pass. The list gives counts, such as “2 of 150 pages blocked”, not points. Heralded checks each page and redirect against that address’s robots.txt before requesting it. If robots.txt refuses Heralded’s reader, Heralded uses the crawl’s own facts to report an error status, a refusal of citation crawlers or a robots meta directive. Otherwise, the server page stays not measured because it was not read. Unreadable robots.txt also leaves access not measured. Redirects are followed within your site’s host and its www. spelling, up to five hops. Ambiguous robots response headers leave a page not measured. This happens when a comma-separated directive could apply to all crawlers or only to the bot named before it, and those readings disagree. An HTTP 429 response is a rate limit and stays not measured after bounded retries. A bot challenge is detected before its robots directives, so a challenge page’s noindex is not reported as your page’s own directive.

The checklist in the Snapshot report

The report’s verdict counts how many of the checks below are not met, and reads “Nothing to fix” only when every check is met. The line under it says whether any page is blocked from AI engines and how many pages were checked. Under the verdict, the report groups the checks into page access, homepage content and structure, and speed, and marks each one met, not met or not measured, with what Heralded measured and the target. A blocked page and its fix sit under the page check it fails. The checks are:
  • the four page checks above, each over every page it read, such as “146/150 pages pass; 4 could not be read”. A page check with no failing page is met, unless most pages could not be read. The page checks read AI crawlers allowed, Page not hidden from AI engines, Page loads for crawlers and Text without JavaScript;
  • the five checks behind the report’s technical moves: Citation crawlers allowed, Main text visible without JavaScript, and three about your homepage’s structured data, Company details for AI (structured data), Company and website described for AI (structured data) and Company name, logo and profiles (structured data).
Structured data, also called JSON-LD, is a block of code in your page that tells AI engines who you are and what you offer. Tooltips on the technical checks explain the terms. Main text visible without JavaScript covers every page with both a server read and a browser text count, including pages that tell AI engines not to use them. Error pages and pages that sent a bot challenge don’t count. A page passes when the server sends at least half the text a browser shows. The check earns 5 points times the share of pages that pass, rounded down. Three passing pages out of four earn 3 of 5 points. The homepage counts only when it has both reads. Before those comparisons arrive, the homepage’s character count gives a provisional score. The check’s score can fall when a page needs JavaScript for more than half its text. Heralded groups those pages into one site fix. Its checklist gives the fix followed by every affected page. Heralded also checks the homepage for page speed: Main content appears quickly, Page responds quickly and Layout stays steady while loading. These are known as largest paint, interaction to next paint and layout shift. Each uses what real visitors on phones experienced over the last 28 days, taken from Chrome’s field data for the page, then for your whole site, then from the copy of that record in PageSpeed Insights. A mobile lab test answers only where no visitor data exists, and it measures total blocking time in place of interaction to next paint. Interaction to next paint earns full points at 200 milliseconds or less and partial points up to 500. Total blocking time earns full points at 200 milliseconds or less and partial points up to 600. PageSpeed Insights can take longer than a read waits. The read then shows these checks as not measured and fills them in when Heralded asks again and the answer arrives. The trend of these three checks and of Main text visible without JavaScript starts again when Heralded changes how it measures them. The server-to-browser comparison starts a new trend; stored reports keep their original results. Heralded also checks HTTPS and HSTS, the sitemap, the canonical URL, Open Graph tags, title and meta description, and domain authority. When one of these or of the three speed checks is not met, it shows with the checks above; the rest sit under Also checked, collapsed on the page and open in the PDF. The verdict counts every check, collapsed or not. The sitemap check fetches every Sitemap: line in your robots.txt. It tries /sitemap.xml only when robots.txt declares none, and reports a missing sitemap only when none of the files it checked supplies a usable sitemap. It accepts gzip-compressed XML and the sitemap protocol limits of 50 MB uncompressed and 50,000 entries per file. Sitemaps explains the inventory read. The Company and website described for AI (structured data) check is met once the homepage’s structured data carries Organization and WebSite. A company description that lacks a name, URL, logo or two social profiles (sameAs in the code) shows what it lacks and keeps the points for the parts it has. Only a homepage with no company description reads as having none. The Snapshot report’s Priority actions link each technical move to the check it would fix. Its Audit section starts a separate Base website audit. Audit reports with no Website assessment show a retired report notice.

Full audit exports

Once your Full audit is delivered, the account that ran it can export its pages, links and findings as CSV, and generate a PDF. Quick and Base audits do not include these exports. The files read the results saved with that audit, including links across every language it read. They do not change when your site changes. The CSV reads are GET /api/scans/{id}/audit/pages.csv, GET /api/scans/{id}/audit/links.csv and GET /api/scans/{id}/audit/findings.csv. Pages include category, language, access, status and findings count. Findings include one row per affected page, their measurement state, share and points off the Website and area scores. Site-wide findings and findings with no affected pages have an empty page cell. A cell that a spreadsheet could interpret as a formula is written as text. POST /api/scans/{id}/generate-pdf returns generating while the PDF renders, or ready with pdf_url once it is available. You can retry a failed render. A delivered Full audit can render without waiting for Snapshot sections. With the WordPress Connector, Heralded can fix robots.txt rules for AI crawlers itself; see Actions. When verified AI crawlers are refused by your CDN or firewall, one action lists every affected page, with steps for the service that recorded the refusals. This work is separate from a robots.txt fix. llms.txt isn’t a check here, and Heralded files no action for it.