Run SEO & AEO Checks in CI with GitHub Actions

Most SEO problems ship in a normal pull request: a title gets shortened, a noindex sneaks in, structured data breaks, an og:image goes missing. You find out weeks later, in a ranking report. The fix is to check the page the same way you check your tests, on every push.

Rank Prompt exposes a public page-audit API (POST /v1/page-audit) that scores a single page against SEO and AEO (Answer Engine Optimization) best practices and returns concrete, ordered fixes. Wire it into GitHub Actions and each PR gets a score and a fix list as a comment, and you can fail the build when a page drops below a threshold.

What gets checked

Each audit returns a 0-100 score, a pass/fail flag, and recommendations across:

• Meta - <title> and meta description length, image alt-text coverage.
• Headings - exactly one <h1>.
• Structured data - Schema.org JSON-LD presence and validity.
• Social - Open Graph and Twitter Card tags.
• Indexability - canonical tag, robots.txt, XML sitemap, and accidental noindex.
• AEO - AI-bot access in robots.txt, /llms.txt, and content freshness (dateModified).
• Security - HTTPS with a valid certificate.

The fastest setup: let your AI do it

If you use Claude Code, Cursor, or any coding agent, you don’t have to hand-write the workflow. In the Rank Prompt app, open Developers -> SEO in CI and copy the setup prompt. Paste it into your AI inside your website repo and it downloads the workflow, sets your page URLs, and opens a PR:

Set up the Rank Prompt SEO/AEO check GitHub Action in this repo:
download the workflow into .github/workflows/rankprompt-seo-check.yml,
set AUDIT_URLS to my key pages, and tell me to add the RANKPROMPT_API_KEY secret.

Or add the workflow yourself

name: Rank Prompt SEO Check
on:
  pull_request:
    branches: [main]
permissions:
  contents: read
  pull-requests: write
jobs:
  seo-check:
    runs-on: ubuntu-latest
    env:
      AUDIT_URLS: "https://example.com https://example.com/pricing"
      FAIL_BELOW: "70"
      RANKPROMPT_API_BASE: "https://api.rankprompt.com"
    steps:
      - name: Audit pages
        env:
          RANKPROMPT_API_KEY: ${{ secrets.RANKPROMPT_API_KEY }}
        run: |
          urls=$(printf '%s' "$AUDIT_URLS" | jq -R 'split(" ")')
          curl -sS -X POST "$RANKPROMPT_API_BASE/v1/page-audit" \
            -H "Authorization: Bearer $RANKPROMPT_API_KEY" \
            -H "Content-Type: application/json" \
            -H "Idempotency-Key: gh-${GITHUB_RUN_ID}" \
            -d "$(jq -n --argjson urls "$urls" '{urls: $urls}')"

Setup is three steps: mint an API key with the read:audit scope, add it as a repository secret named RANKPROMPT_API_KEY, and list your pages in AUDIT_URLS. Each audited page costs 1 credit, and a page that can’t be fetched is refunded automatically.

Auditing a whole site at once

The same endpoint takes one url, a list of urls, or a sitemap. Pass a sitemap (or several URLs) and it audits them in a single call, fetching the domain-level checks (robots, sitemap, llms.txt, SSL) once instead of per page.

curl -sS -X POST "https://api.rankprompt.com/v1/page-audit" \
  -H "Authorization: Bearer $RANKPROMPT_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: site-audit-1" \
  -d '{"sitemap": "https://example.com/sitemap.xml", "limit": 10}'

You only pay for pages that audit successfully, and the response includes credits_charged so the cost is never a surprise.

Why AEO belongs in CI

Traditional SEO linters check tags. The Rank Prompt audit also checks the things that decide whether AI assistants can read and cite you: AI-bot access in robots.txt, an /llms.txt manifest, valid structured data, and dated, fresh content. Catching those in the PR that introduces them is far cheaper than discovering, a month later, that ChatGPT stopped mentioning your brand.

Add the check once, and every future change to your site is measured against it before it ships.