Scoring a site for answer engines: brevio-core aeo

In Plain English (For Anyone)	Technical Contract (For Engineers)
What it does: Grades your built website HTML, robots.txt, and llms.txt so AI search engines (Perplexity, ChatGPT, Claude) can discover and cite your pages accurately. When to use it: On your static build folder (out/, dist/, .next/) locally or in CI.	Command: brevio-core aeo <folder> [--min SCORE] Inputs: Built .html files, robots.txt, llms.txt, skill.md Exit Codes: The exit code is 0 when nothing is found, 1 when not.

brevio-core aeo scores how well the pages of a built site can be found, read and cited by answer engines (AEO). It is a separate command from the pull-request gate and reads a folder of built HTML pages:

brevio-core aeo out/                      # a scorecard of every page, worst first
brevio-core aeo out/ --min 75             # exit 1 when a page scores below 75
brevio-core aeo out/ --format markdown    # a table for a job summary or a pull request comment
brevio-core aeo out/pricing.html --format json

What it reads

• Every .html page under the folder, hidden folders and node_modules left out; or the one page you name.
• At the root of the folder (for one page, its own folder): robots.txt, llms.txt, skill.md and .well-known/agent-card.json. A missing file counts as missing, not as an error.
• A Markdown copy beside a page (page.md or page.html.md), which counts as the page’s Markdown mirror.

It reads local files only

aeo does not fetch anything. Given a URL (an address with a scheme such as https:, or a bare host name such as www.example.com) it stops with exit code 2 and says what to do instead:

1. Save the built page: your site’s build output folder is best; a browser’s “Save page as” works for one page.
2. Put the site’s robots.txt and llms.txt beside it.
3. Run brevio-core aeo on that folder.

To score a live URL without saving it, use the hosted check on brevio.pro.

The scorecard

Each page gets a score out of 100 and a grade from A to F, from five parts:

Part	What it looks at
Discovery	whether robots.txt shuts out the AI crawlers, llms.txt and its per-page token counts, a canonical URL
Structure	text a crawler sees without running scripts, a heading tree, a table, JSON-LD structured data, enough text to quote
Token economics	how much of the page is markup, the page’s size, a Markdown copy
Capability signals	skill.md, an agent card or an MCP descriptor
A link for people	a one-click link a human reader can follow from a cited answer

Under each page come its gaps, the first ones first. The text and Markdown reports start with what the site files say; --format json adds each part’s score.

Exit codes

Code	When
0	every page scores --min or more (the default --min is 0: report only)
1	at least one page scores below --min
2	a usage error: a URL, a missing path, a folder with no .html page, or a bad --min

In CI

Run it after the build step, on the build output:

      - run: npm run build
      - run: pip install brevio-core==0.9.1
      - run: brevio-core aeo out/ --min 75 --format markdown >> "$GITHUB_STEP_SUMMARY"