The rules of Agentic SEO Check.

What to change: Keep exactly one <title>. Browsers and crawlers may pick different ones when several exist.

What to change: Shorten the title to roughly 60 characters or fewer, keeping the most important words first.

What to change: Replace the placeholder with a title that describes this page. In frameworks, set it per route (metadata export, <Head>, useHead, <svelte:head>, etc.).

What to change: Add a unique meta description (about 70-160 characters) that summarises the page and invites the click. Search engines may show it as the snippet.

What to change: Keep a single meta description; remove or merge the others.

What to change: Trim the description to about 160 characters, front-loading the key message.

What to change: Add one <h1> that states the main topic of the page. It is usually the visible page heading.

What to change: Put the page heading text inside the <h1>. If it wraps only an image, give the image descriptive alt text.

What to change: Remove empty heading tags or fill them with text. Empty headings confuse outline parsers and screen readers.

What to change: Add a self-referencing canonical with the absolute preferred URL (scheme, host, trailing slash) so duplicate URLs (query strings, www/non-www) consolidate to it.

What to change: Use an absolute URL in the canonical link. Relative canonicals are error-prone and some crawlers resolve them unexpectedly.

What to change: Point the canonical at the https URL.

What to change: Give each page of the series a canonical that names its own URL (?page=2 names ?page=2), and link the pages to each other with plain <a href> links. Google advises against making the first page the canonical of the series.

What to change: If this page should appear in search results, remove the noindex directive. Preview/staging hosts often add X-Robots-Tag: noindex at the platform level; verify on production. If the page is intentionally private (login, thank-you, internal search), this is correct.

What to change: Write each directive the way Google documents it, separated by commas: content="noindex, nofollow". A misspelled noindex is ignored, so the page stays indexable.

What to change: Remove unavailable_after if the page should show in search again; a template or plugin often leaves an old date in place. To have a page drop out later, write the date in ISO 8601 (2026-12-31), RFC 822 or RFC 850.

What to change: Redirect on the server with a 301 (or 308) to the new URL. If the server cannot redirect, set the meta refresh to 0 seconds, which Google reads as a permanent redirect.

What to change: Set <html lang="xx"> (for example "en", "nl", "en-GB") so search engines and screen readers know the content language.

What to change: Use a BCP 47 tag such as "en", "en-US" or "pt-BR".

What to change: Add <meta name="viewport" content="width=device-width, initial-scale=1"> to <head>. Google indexes the mobile version of pages first.

What to change: Move these elements into <head>. Google accepts rel=canonical only in <head>. A framework that renders them in a page component usually has a head API for them (metadata export, <Head>, useHead, <svelte:head>).

What to change: Load every resource over https: change http:// URLs to https:// (or make them relative), and host or replace any resource that has no https version. A Content-Security-Policy of upgrade-insecure-requests upgrades the rest.

What to change: Add og:title, og:description, og:image (absolute URL, at least 1200x630), og:url and og:type so shares on social platforms and chat apps render a rich preview.

What to change: Use a full https:// URL for og:image; most platforms ignore relative image paths.

What to change: Use one of the supported card types, usually summary_large_image.

What to change: Give every <img> an alt attribute: a short description for meaningful images, alt="" for purely decorative ones. Alt text is used for image search and accessibility.

What to change: Link to related pages and the main navigation with plain <a href> elements.

What to change: Add descriptive link text, or aria-label for icon links. Empty links tell crawlers nothing about the target page.

What to change: Route with the History API instead of fragments, so each view has a path (/products) the server answers, and link to those paths.

What to change: Update internal links to https:// (or use relative URLs) to avoid an extra redirect on every click and crawl.

What to change: Keep sessions in a cookie, not in the URL. In Java servlet containers, set the session tracking mode to COOKIE; in PHP, set session.use_trans_sid=0 and session.use_only_cookies=1.

What to change: Turn on gzip or Brotli compression for text/html (and CSS, JavaScript, JSON, SVG) in the web server, CDN or hosting settings: gzip on in nginx, mod_deflate in Apache, compression() middleware in Express.

What to change: Make the largest element of the first screen show sooner: never lazy-load it, give its image fetchpriority="high", defer the scripts in <head>, and speed up the server's first byte (cache the HTML at the edge). The Chrome UX Report covers 28 days, so a fix shows there over the next four weeks; check it now with --lighthouse on the local tool, or PageSpeed Insights, which also names the element.

Runs with --field-data (fieldData), on the command and both MCP servers, and in this site's audits: it reads the Chrome UX Report's real-user data.

What to change: Shorten the work a tap or a key starts: split long JavaScript tasks (await scheduler.yield()), defer scripts in <head>, load other sites' scripts after the page, and keep the viewport meta tag so phones answer a tap at once. The Chrome UX Report covers 28 days, so a fix shows there over the next four weeks; check it now with --lighthouse on the local tool, or PageSpeed Insights, which also names the element.

Runs with --field-data (fieldData), on the command and both MCP servers, and in this site's audits: it reads the Chrome UX Report's real-user data.

What to change: Give every image width and height, reserve the space of what a script inserts (banners, ads, embeds), and load web fonts with font-display: optional. The Chrome UX Report covers 28 days, so a fix shows there over the next four weeks; check it now with --lighthouse on the local tool, or PageSpeed Insights, which also names the element.

Runs with --field-data (fieldData), on the command and both MCP servers, and in this site's audits: it reads the Chrome UX Report's real-user data.

What to change: Preload the LCP image/font, serve it from a CDN in a modern format, inline critical CSS, and cut server response time.

Runs with --lighthouse, on the command or the local server.

What to change: Reserve space for images, ads and embeds with width/height or aspect-ratio; avoid inserting content above existing content; use font-display: optional or size-adjusted fallbacks.

Runs with --lighthouse, on the command or the local server.

What to change: Split long JavaScript tasks, defer non-critical scripts, remove unused JavaScript, and move work off the main thread.

Runs with --lighthouse, on the command or the local server.

What to change: Cache HTML at the edge, avoid slow data fetches during render, and use a nearby region/CDN.

Runs with --lighthouse, on the command or the local server.

What to change: Work through the opportunities listed in the performance result, biggest savings first.

Runs with --lighthouse, on the command or the local server.

What to change: Follow the Lighthouse audit description in the finding.

Runs with --lighthouse, on the command or the local server.

What to change: Use codes like "en", "en-gb", "nl-nl", "zh-hant" or "x-default". Country codes alone ("uk", "eu") are invalid.

What to change: Use fully qualified URLs (https://example.com/nl/) in every hreflang link.

What to change: Each language/region code may map to exactly one URL per page.

What to change: Every page in an hreflang cluster must list itself as well as its alternates, otherwise the annotations may be ignored.

What to change: Give each language version a canonical that names its own URL, and let hreflang link the versions. Google asks for a canonical in the same language as the page.

What to change: Set "@context": "https://schema.org" on every top-level JSON-LD object.

What to change: Add "@type" (e.g. "Article") to every structured data object.

What to change: Add the missing properties listed in evidence. Validate with https://search.google.com/test/rich-results after fixing.

What to change: Use ISO 8601 dates such as 2026-03-14 or 2026-03-14T09:30:00+01:00.

What to change: Write prices as plain numbers with a dot ("19.99") and the currency in priceCurrency ("EUR"); set bestRating when ratings are not out of 5; number breadcrumb positions 1, 2, 3 and give every crumb but the last an item URL; write availability, itemCondition, eventStatus and eventAttendanceMode as schema.org values ("https://schema.org/InStock"), and employmentType as FULL_TIME, PART_TIME, CONTRACTOR, ... in capitals.

What to change: Give the object a hasPart of type WebPageElement with "isAccessibleForFree": false and a cssSelector that is a single class (".paywall"), and give the element that holds the paywalled text that class in the HTML.

What to change: If the page is rendered client-side, use server-side rendering or static generation so the HTML carries the content. Re-run with render enabled to compare the rendered DOM.

What to change: Add substantive, useful content, or set noindex if the page is a utility page (login, cart, thank-you). Thin pages rarely rank and can drag down site quality signals.

What to change: Include the target keyword (or a close variant) near the start of the title.

Runs when the audit is given a keyword.

What to change: Use the target keyword in the H1 so the main heading matches the search intent.

Runs when the audit is given a keyword.

What to change: Use the keyword naturally in the body copy, ideally in the first paragraph.

Runs when the audit is given a keyword.

What to change: Collapse the chain so the first URL redirects straight to the final destination. Each hop costs crawl budget and loses some signal.

What to change: Reduce server response time: cache the HTML, avoid blocking data fetches during render, use a CDN. Google recommends TTFB under 800 ms.

What to change: Serve pages with Content-Type: text/html; charset=utf-8.

What to change: Serve the site over https and redirect http to https. HTTPS is a (light) ranking signal and required for many browser features.

What to change: Render this in the server HTML (SSR/SSG/prerender). Googlebot queues JavaScript rendering separately and other crawlers and link previews never run it.

Runs with --render, on the command or the local server.

What to change: Render this in the server HTML (SSR/SSG/prerender). Googlebot queues JavaScript rendering separately and other crawlers and link previews never run it.

Runs with --render, on the command or the local server.

What to change: Render this in the server HTML (SSR/SSG/prerender). Googlebot queues JavaScript rendering separately and other crawlers and link previews never run it.

Runs with --render, on the command or the local server.

What to change: Emit the final canonical in the server HTML and do not modify it client-side.

Runs with --render, on the command or the local server.

What to change: Remove noindex from the server HTML if the page should be indexed.

Runs with --render, on the command or the local server.

What to change: Render this in the server HTML (SSR/SSG/prerender). Googlebot queues JavaScript rendering separately and other crawlers and link previews never run it.

Runs with --render, on the command or the local server.

What to change: Render this in the server HTML (SSR/SSG/prerender). Googlebot queues JavaScript rendering separately and other crawlers and link previews never run it.

Runs with --render, on the command or the local server.

What to change: Render this in the server HTML (SSR/SSG/prerender). Googlebot queues JavaScript rendering separately and other crawlers and link previews never run it. At minimum, render the main navigation and content links server-side.

Runs with --render, on the command or the local server.

What to change: Add /robots.txt. At minimum: "User-agent: *" with no disallow rules, plus a "Sitemap:" line.

Runs on a crawl or a build directory.

What to change: Make /robots.txt return 200 (or 404 if you have none).

Runs on a crawl or a build directory.

What to change: Allow the paths in robots.txt, or stop linking to them and remove them from the sitemap if they should stay hidden.

Runs on a crawl or a build directory.

What to change: Shorten robots.txt: group paths under shared prefixes, use wildcards (Disallow: /*?sort=) instead of listing URLs one by one, and remove rules for paths that no longer exist.

Runs on a crawl or a build directory.

What to change: Remove noindex: and nofollow: lines and use a robots meta tag or X-Robots-Tag header on the pages instead. Crawl-delay may stay for Bing and other crawlers; Google sets its crawl rate itself.

Runs on a crawl or a build directory.

What to change: Start every Allow and Disallow path with / (Disallow: /wp-admin/, not wp-admin/ or https://example.com/wp-admin/), put every rule under a User-agent line, and write each Sitemap line as a full URL (Sitemap: https://example.com/sitemap.xml).

Runs on a crawl or a build directory.

What to change: Allow the CSS and JavaScript files pages need (Allow: /assets/, or remove the Disallow line for that folder). Keep blocking only what pages do not need to render.

Runs on a crawl or a build directory.

What to change: Generate a sitemap.xml listing canonical, indexable URLs and reference it from robots.txt.

Runs on a crawl or a build directory.

What to change: List every sitemap file directly in one sitemap index (up to 50,000), or submit each index separately in Search Console.

Runs on a crawl or a build directory.

What to change: Write every <loc> as a full URL with protocol and host (https://www.example.com/about/, not /about/), under 2,048 characters, generated from the production site URL.

Runs on a crawl or a build directory.

What to change: Write <lastmod> as a date (2026-11-09) or a date and time with its time zone (2026-11-09T10:00:00+01:00).

Runs on a crawl or a build directory.

What to change: List only canonical URLs in the sitemap.

Runs on a crawl or a build directory.

What to change: Replace redirecting URLs in the sitemap with their final destination.

Runs on a crawl or a build directory.

What to change: Update the links to the final URL to save a round trip for users and crawlers.

Runs on a crawl or a build directory.

What to change: Point each redirect straight at its final destination.

Runs on a crawl or a build directory.

What to change: Give each page a unique title that describes that page (templates: include the item name, category and brand).

Runs on a crawl or a build directory.

What to change: Write a unique description per page, or generate it from page-specific data.

Runs on a crawl or a build directory.

What to change: Consolidate duplicates with a canonical to one URL, 301 redirect, or make the content distinct.

Runs on a crawl or a build directory.

What to change: Return a real 404 (or 410) status for unknown routes; otherwise search engines index error pages and waste crawl budget.

Runs on a crawl or a build directory.

What to change: Add a permanent (301/308) redirect from http to https at the server or CDN.

Runs on a crawl or a build directory.

What to change: Redirect the alternate host (www or bare domain) to the canonical host with a 301/308.

Runs on a crawl or a build directory.

What to change: Every page in an hreflang cluster must list all alternates including itself; mismatches invalidate the annotations.

Runs on a crawl or a build directory.

What to change: Fix or remove the dead outbound links; they hurt user trust and signal an unmaintained page.

Runs on a crawl or a build directory.

What to change: Write the URL the redirect ends at in the canonical: the same scheme, host, trailing slash and case the server answers with.

Runs on a crawl or a build directory.

What to change: Remove noindex from the canonical page, or point the canonical somewhere indexable. If both pages should stay out of search results, use noindex on both and drop the canonical.

Runs on a crawl or a build directory.

What to change: Point every page's canonical straight at the final preferred URL.

Runs on a crawl or a build directory.

What to change: List only the canonical, indexable URL of each language version in hreflang: the final URL, not one that redirects.

Runs on a crawl or a build directory.

What to change: Redirect (301) the other spelling to the page's URL in the server, framework or CDN (trailingSlash in Next.js, Nuxt and Astro configs; a lowercase rewrite on case-insensitive servers), or at least give every spelling a canonical to the one URL.

Runs on a crawl or a build directory.

What to change: Give the parameter versions a canonical to the URL without the parameter (or the one you want indexed), and link to that URL. Parameters that change nothing visible (session, sort order of one item, view=) should not reach links at all.

Runs on a crawl or a build directory.

What to change: Fix the image URLs listed in evidence, or upload the missing files. Each entry names the pages that show the image.

Runs on a crawl or a build directory.

What to change: Remove noindex if answer engines should quote this page. Keep it on pages that should stay private (login, cart, internal search); preview hosts often add it with X-Robots-Tag, so check production first.

What to change: Remove nosnippet or max-snippet:0 if answer engines should quote this page. To hide one part only, wrap that part in an element with data-nosnippet.

What to change: Remove noarchive if the page should appear in Copilot answers. Paywalled pages can use nocache instead, which lets Copilot cite the URL, title and snippet. Google ignores noarchive.

What to change: Render the content on the server (SSR or static generation) so the HTML carries it. `agentic-seo-check audit <url> --render` compares the HTML with the rendered page.

What to change: Mark up only what the page shows: render every question, answer and step from the same data as the JSON-LD, or remove the items that are not on the page. Google's structured data guidelines forbid marking up content readers cannot see. If the page adds them with JavaScript, render them on the server too: most AI crawlers do not run JavaScript.

What to change: Add author as a Person (or Organization) with name and url, the url pointing to an author page. Answer engines weigh who stands behind the content.

What to change: Add datePublished and dateModified in ISO 8601 (2026-03-14T09:30:00+01:00), and update dateModified when the content changes materially.

What to change: Make /robots.txt answer 200, or 404 when there is none. A 5xx or a timeout stops well-behaved crawlers, AI crawlers included.

What to change: Allow the bots listed in evidence if the site should be cited in those answer engines: add a group for each with `Allow: /`, or remove the Disallow lines that match them. Keep a block that is deliberate, and disable this rule.

What to change: Set ai-input=yes and search=yes in the Content-Signal line if answer engines should use the content; ai-train=no can stay to opt out of training. Example: `Content-Signal: search=yes, ai-input=yes, ai-train=no`.

What to change: Start the file with `# Site name`, then an optional `> one-line summary`, then H2 sections with links written as `- [Title](https://...): note`.

What to change: Allow these bots at the CDN, firewall or bot-protection layer (for example the CDN's AI crawler settings or a WAF rule). The probe sent the bots' User-Agent from this machine, not from the bots' own IP ranges, and a CDN that verifies bots may let the real ones through: confirm in the CDN's bot logs before changing rules.

Runs with --ai-bots, on the command or the local server.