Visibility tracking
The AEO Toolkit answers “is this site built to be cited?” It crawls a site, scores it against the 44 technical SEO and AEO checks (plus a separate 14-signal E-E-A-T scan), and renders a scorecard. That is a point-in-time, on-page diagnostic.
It does not answer the question a site owner cares about most: “is it working? Are the AI engines now citing me for the questions my customers ask, and is that improving?” That is an off-page, over-time outcome, and it lives in a separate companion system: the AEO Visibility Tracker.
This page explains what the tracker is, why it lives outside this repo, and how the two systems fit together.
Where the code lives: the tracker is closed-source hosted tooling and is not in this repo. This repo (the toolkit) stays a clean, public, Apache-2.0 audit suite.
Two halves of one job
| AEO Toolkit (this repo) | Visibility Tracker (hosted service) | |
|---|---|---|
| Question | Is the site built to be cited? | Is the site actually being cited, over time? |
| Layer | On-page (the site itself) | Off-page (the AI engines’ answers) |
| Method | Crawl → 44-check score → PDF | Ask AI engines the questions customers ask → record cited / rank |
| Cadence | Point-in-time (per audit) | Recurring (weekly time series) |
| Output | Scorecard + prioritized fixes | cited 0% → 67%, avg rank 4.0 → 1.5 deltas |
| Role | Baseline + the fix list | Proof the fixes moved the needle |
The toolkit audit sets the baseline score and the list of fixes; the tracker shows, week over week, whether that work changed what the AI engines actually say.
How the tracker works
For each tracked site, question and engine, it asks an AI search engine the questions that site’s customers type (“best HVAC in London Ontario”), then analyzes the answer:
- cited: did the business appear in the answer text or in the engine’s cited sources?
- rank: its 1-based position among named competitors in the prose (
nullif only matched via a source URL, because a footnote citation is not a recommendation ranking). - competitors_cited: which tracked competitors won the slot instead.
- sources / excerpt: evidence, meaning the citation URLs and a snippet around the mention.
Each check is one timestamped row. Diffing the history yields the before and after.
Engines: Perplexity, OpenAI (with web search) and Gemini, called through one gateway so results stay comparable across engines. Each row records which engine answered.
Cadence: a weekly scheduled run, plus ad-hoc runs for backfills and reports.
The join point: audit_score
The tracker’s schema reserves an audit_score column to bridge the two systems. A planned
enhancement runs this toolkit’s @advance-labs/scoring on each tracked site per run and logs the
on-page score next to the off-page citation result, so one table tells the whole story:
“The AEO score went from F (38) to A (91); over the same 6 weeks the ChatGPT citation rate went 0% → 67% and the average rank 4.0 → 1.5.”
The audit explains why; the tracker proves that.
See also
- This repo: Architecture and
packages/scoring, the audit engine the baseline is built on. - The hosted service: advancelabs.dev/services/ai-visibility.
AEO Toolkit is an open-source instrument by Advance Labs.