← Home

SEOAICo

What AI says about a business, kept on the record.

I designed and wrote the software for this platform. It is a finished product that a business or an agency can buy and run.

It asks ChatGPT and Perplexity the same questions on a schedule and saves every answer in full. Then it sorts each one: linked, named without a link, or left out. Those are three different situations, so I kept them apart and didn’t blend them into one score.

A second part reads the website itself, runs twenty checks across six areas, and gives a score out of a hundred. The single-page scan on the live site uses that same scoring.

It’s running today and collecting data. What I’d bring to an in-house team is what went into it: how search works, how to measure it, and how to explain the results to the person who owns the business.

What it’s made with

What starts workSchedulerruns on a fixed cadenceA submitted URLone page, scanned nowBackground job queuea database table, worked by supervisorPipeline oneCitation probesquestions × services × placesChatGPT · Perplexitythe whole reply is storedcited · mentioned · absentPipeline twoPage scanfetch, after a public-address check20 checks, 6 categoriesdeterministic, every timea score out of 100SQLiteone fileFilamentoperator console + reports
How the pieces are wired: one queue, two pipelines that share nothing but the database, and one console reading from it.

What it looks like

A readout strip naming the scanned domain, a score of 90, a stable state and the time of the last evaluation, above three tiles reading 18 of 20 checks passed, strongest area technical accessibility 10 out of 10, and weakest area entity definition strength 5 out of 15.
The readout for a scanned domain: the score, and the strongest and weakest signal areas behind it.
A scan result panel for playplayai.com showing a score of 90 drawn inside a ring labelled AI visibility score, a readiness heading, and the single highest-impact finding: add clear descriptions of what the business offers using natural language.
A real scan of playplayai.com, run against a local build of the platform while this page was written.
Seven cards, one per scored category, each giving points earned out of points available, how many checks passed, a progress bar, and a disclosure that opens the checks behind the score.
Each category opens to the individual checks, so a score can be taken apart rather than taken on faith.

Visit the live site → (opens in a new tab)

You might also like

  • AI Citation Benchmark

    How often does AI cite a brand? It asks AI assistants a set of real questions and sorts every answer into three groups: cited, named but not linked, or left out.

  • Node Bloom

    Watch an AI answer take shape. A live drawing of how one source spreads through AI answers.

← See everything