SEO and AEO scan: rank and get cited

The SEO and AEO scan crawls your whole site and gives you two separate scores: an SEO score for how well your pages can rank on Google and Bing, and an AEO score for how well AI engines like ChatGPT, Claude, and Perplexity can read and cite them. Every issue it finds comes with a copy-paste AI fix prompt you drop into Claude, ChatGPT, or Cursor, so you don’t need an SEO consultant to act on it. It’s on every plan.

Two scores, and why both matter in 2026

Search has split into two surfaces, and a site can do well on one while being invisible on the other.

• SEO (search engine optimization) is the old, still-important game: ranking in Google’s and Bing’s blue links. Google renders JavaScript, reads your titles and headings, and ranks accordingly.
• AEO (answer engine optimization) is getting read and cited inside an AI answer: ChatGPT, Claude, Perplexity, Google AI Overviews, Copilot. These engines synthesize an answer instead of returning a list, so the goal is to be the source they quote, not just a link they rank.

The two diverge in a way that catches people out. AI crawlers (GPTBot, ClaudeBot, PerplexityBot) generally do not run JavaScript, so a page whose content loads with JavaScript can rank fine on Google (which renders it) and still be a blank shell to an AI engine (which doesn’t). That’s exactly the kind of gap the AEO score surfaces. As more searches end inside an AI answer, being inside the answer matters as much as ranking under it.

A note on honesty: AEO is a fast-moving area. The scan checks the technical things that are clearly under your control and clearly help AI engines read you (content in the HTML, schema, clear headings, crawler access). It does not promise a specific ranking or citation. It also flags an llms.txt as a low-priority, emerging-standard item, not a must-have, because no major AI provider has confirmed it reads llms.txt in production.

What the scan needs

• An Amabrik account with the site added.
• A verified domain. The scan only runs against a domain you’ve proven you own. If the domain isn’t verified, the scan returns an error and prompts you to verify first.
• That’s it. The scan needs no API key and never queries an AI. It reads your site’s own robots.txt, raw HTML, and markup. It only ever reports; it never changes anything on your live site.

Verify your domain

The scan is gated on a verified domain so nobody can scan a site they don’t control.

1. In the dashboard, add your site and open the SEO and AEO page (or Settings, where domains live).
2. Amabrik shows you a DNS TXT record to add at your domain’s DNS provider.
3. Add the TXT record, then click Verify. Amabrik checks your DNS for the record. On some providers this is a one-click flow; on others you paste the record in by hand.
4. Once the record is found, the domain is marked verified and the Scan button unlocks.

DNS changes can take a few minutes to propagate. If verification doesn’t pass right away, wait a moment and try again.

Run a scan

With a verified domain, open the SEO and AEO page and click Scan. Amabrik crawls your site starting from the homepage, follows your internal links, and analyzes every HTML page it finds (it skips assets, non-HTML responses, and pages that return an error).

• A small site finishes in seconds and shows the report inline.
• A larger site continues crawling in the background and the page polls for the result, so a site of any size gets fully scanned without timing out.

How many pages get crawled depends on your plan:

During the free trial the scan checks your homepage only and shows the two scores and the issues, but the copy-paste fix prompts are unlocked once you’re on a paid plan. If a scan hits your plan’s page cap, the report tells you it was truncated and how many pages it covered.

What the SEO score checks

The SEO score starts at 100 and drops by the severity of each issue found. These are the real on-page and site-level checks, run per page where noted.

Per page:

• Title tag. Flags a missing <title> (high), a title over 60 characters (Google truncates past about 60), or a very short title under 15 characters.
• Meta description. Flags a missing description (Google then writes its own snippet) or one over 160 characters (it gets cut off).
• H1 heading. Flags a page with no <h1>, or with more than one <h1> (which muddies the main topic).
• Canonical link. Flags a page with no <link rel="canonical">, so duplicate or parameter URLs can split your ranking.
• Mobile viewport. Flags a missing viewport meta tag, which Google ranks on.
• Open Graph. Flags missing og:title / og:image, which control how the page looks when shared and in some previews.
• Image alt text. When a page has at least three images and half or more have no alt, it flags them. Alt text helps image search, screen readers, and how AI reads the page.
• Unintended noindex. Flags a page set to noindex so you can confirm each one is intentional (correct for login or cart pages, a costly mistake on a content page).
• Thin content. Flags a page with very little readable text (under about 600 characters of visible text), which rarely ranks.

Site-level (checked once):

• robots.txt. Flags a missing robots.txt, or one that blocks the whole site with a blanket Disallow: / (high, since it can drop you from Google entirely).
• Sitemap. Flags when there’s no sitemap.xml and none referenced in robots.txt.
• HTTPS. Flags a site served over HTTP instead of HTTPS (high).
• Language attribute. Flags an <html> tag with no lang attribute.
• Broken internal links. While crawling, any internal link that returns a 404 is recorded, along with the page it was linked from, and reported as one issue. Google follows the same links, so a dead one wastes crawl budget and hurts the page that links to it.

What the AEO score checks

The AEO score is scored the same way, from 100 down by severity, but on AI-readiness. These checks read your raw HTML and robots.txt; they never call an AI.

Per page:

• Content readable by AI (the big one). If a page’s main content is rendered with JavaScript (the raw HTML is an app shell with no <h1> and a framework signal like a Next.js, Nuxt, React, or Angular root), it’s flagged high: AI crawlers don’t run JavaScript, so they see only your navigation. Google renders JS and still ranks you, which is why this is an AEO issue, not a Google one. The scan also notes when your H1 specifically is added by JavaScript, so an AI engine never learns what the page is about. (On these pages the scan skips the H1, headings, and thin-content checks so it doesn’t report false positives for content that’s there but JS-rendered.)
• Structured data (schema.org). Flags a page with no JSON-LD (application/ld+json) so you can add Article, FAQPage, HowTo, or Organization markup that lets an engine lift facts cleanly. The check detects schema whether it’s a plain <script> tag or serialized in a framework’s hydration payload, so it won’t tell you to add schema you already have.
• Subheadings to lift answers from. Flags a page with no <h2> or <h3>. AI engines pull answers from clear headed sections; a wall of text is hard to quote.

Site-level (checked once):

• AI crawler access. The scan parses your robots.txt and checks each of seven AI engines against the crawler tokens it uses: ChatGPT (GPTBot, OAI-SearchBot), Claude (ClaudeBot, anthropic-ai), Perplexity (PerplexityBot), Google AI (Google-Extended), Copilot (Bingbot), Meta AI (meta-externalagent, FacebookBot), and Mistral (MistralAI-User). It reports only the engines that are blocked (so a clean report stays quiet), because a blocked crawler means that engine can never read or recommend you.
• llms.txt. Flags a missing /llms.txt as a low-priority, emerging-standard item: a plain-text map of your key pages that some AI tooling looks for. Treat it as optional, not required.

Read the report and use the AI fix prompt

The report shows your two scores side by side, the number of pages scanned, and the issues grouped by type. An issue that hits many pages (say, a missing meta description on 50 pages) appears once with the count and the affected page paths, not as 50 separate entries, so your score and your to-do list both stay sane.

Each issue gives you three things, in plain English:

1. What’s wrong on the page.
2. Why it matters for ranking or for AI visibility.
3. A copy-paste AI fix prompt.

The fix prompt is a precise, scoped instruction written for your AI assistant. Copy it, paste it into Claude, ChatGPT, or Cursor, and it writes the change for you. The prompt names every affected page and ends with a sweep of the rest of the site, so the fix closes the issue everywhere, not just on one page. The prompts are written to be surgical: they only add or correct tags, and any robots.txt edit allows one crawler rather than rewriting the file, so a fix doesn’t break a working page. Many issues also include a short test after checklist (for example, “disable JavaScript and reload, the main content should still show”) so you can confirm the fix landed.

You review the change before you ship it, like any other edit. The scan itself never touches your live site.

Scheduled scans

You can re-scan automatically so a regression is caught the day it ships, not at your next manual audit. Open the gear (automatic scans and alerts) on the SEO and AEO page and set:

• Frequency: Off, Daily, Weekly, or Monthly. Scheduled scans run in an off-peak hour based on your timezone.
• Email me: only new high-priority issues, after every scan, or never.
• Notification email and an optional webhook (for example, a Slack incoming webhook).

When a scheduled scan runs, Amabrik first checks whether your homepage has changed since the last scan (by hashing it). If nothing changed, it skips the crawl, so you’re not re-scanned for no reason. When it does run and finds issues that weren’t in the previous scan, it emails (and webhooks) you the new ones, each with its fix prompt, based on the alert mode you chose.

FAQ

What is AEO?

AEO is answer engine optimization: getting your content read and cited by AI answer engines like ChatGPT, Claude, Perplexity, and Google AI Overviews, rather than only ranked in Google’s links. The Amabrik AEO score measures the technical things that decide whether those engines can read your page at all: content in the HTML, schema, clear headings, and crawler access.

How is AEO different from SEO?

SEO is about ranking in search engine results. AEO is about being the source an AI answer quotes. They overlap, but they diverge on JavaScript: Google renders it and ranks you, while AI crawlers usually don’t, so a page can score well on SEO and poorly on AEO. Amabrik scores them separately so you see both.

Do I need a verified domain?

Yes. The scan only runs against a domain you’ve verified with a DNS TXT record, so nobody can scan a site they don’t own. The Scan button stays locked until verification passes.

Does a good score guarantee I’ll rank or get cited?

No. The scan fixes the technical and on-page things that hold a page back, but ranking and AI citation depend on content quality, authority, competition, and how each engine weighs signals, which no tool controls. A clean scan removes the blockers; it doesn’t buy a ranking.

Do I need to know SEO or how to code?

No. Every issue is explained in plain English and comes with a copy-paste AI fix prompt. Paste it into Claude, ChatGPT, or Cursor and it writes the change. You review and ship it.

How often should I run a scan?

After every deploy or content change, and at least monthly. A redesign or an AI-assisted edit can quietly break things: a page goes client-rendered, a title gets dropped, a crawler gets blocked. A daily or weekly scheduled scan catches it the day it ships.

Does the scan query ChatGPT or any AI to test my site?

No. It reads your own robots.txt, raw HTML, and markup. It never calls an AI and needs no API key, which is why it’s fast and on every plan. (The fix prompts are for you to paste into your own AI assistant.)

Is the AI fix prompt safe to paste in?

Yes. It’s a scoped instruction that describes the change. The prompts only add or correct tags, and any robots.txt edit is surgical, so they don’t break a working page. You review the diff before shipping, and the scan itself never changes your live site.

Will it work on a no-code or AI-built site?

Yes. Sites built with tools like Lovable, Bolt, v0, Framer, or Webflow often ship client-rendered pages that AI can’t read and skip schema entirely. The scan catches exactly those gaps and hands you the fix prompt to close them.