We can't find the internet
Attempting to reconnect
Something went wrong!
Attempting to reconnect
TabsTX · perception
The press the machines actually read
Every time an AI answer engine recommends a Dallas restaurant, it cites its sources. We keep the receipts. This is the ranked list of publications and individual articles the engines pulled from — which is, in practice, the list of places worth getting written about.
- Citations
- 7824
- Publications
- 1209
- Editorial
- 41
- Answers cited
- 1032
Weighted by citation, not by URL — a five-year-old round-up cited three hundred times shapes three hundred answers. Dated for 0% of 3079 article citations; the rest publish no date we can read, and are left out rather than guessed at.
Nigerian
Directoryhttps://www.ubereats.com/category/irving-tx/nigerian
Questions these citations answered
What a diner was asking when an engine reached for this source.
- Best West African / Nigerian food in DFW? CClaude 1 1
Venues named alongside it
Rooms the engines named in the same answers that cited this source. Co-occurrence, not proof of cause — but it is the shape of the influence.
No resolved venues in the answers that cited this source yet.
Method
We ask the AI answer engines the questions diners actually ask, on a schedule, and store every answer with its citations. The counts above are citation rows — one per source URL an engine attached to an answer — scoped to the most recent complete observation run.
"Editorial" versus "directory" is our judgement call, applied by domain and tuned as the corpus grows. Editorial means a publication with an editor: a place a restaurant can be pitched. Directory means an aggregator, listing site, OTA, or user-generated platform — visible, but not pitchable.
Publication dates and bylines come from each cited page's own structured data — the publisher's JSON-LD, then its OpenGraph tags. We fetch politely, once per second, identify ourselves, and stop asking any site that declines. Where a page declares no date we say so; we never estimate one. The exception is Reddit, which refuses crawlers but encodes a post's position in its own submission counter in the URL — those dates are arithmetic, accurate to roughly a day, and marked approximate wherever they appear.