The category spent three weeks explaining Reddit's citation collapse with a robots.txt change that never happened, and the vendor who collected the data blames a ChatGPT change this radar already carried on 21 August.

Reddit held a steady 3.8% share of ChatGPT Search citations from 18 July through 7 August. The 14 to 17 August average was 0.5%, which Promptwatch reports as an 86.4% relative drop. Google's AI Overviews fell 11.3% and AI Mode 30.5% over the same period. Promptwatch caveats its own number: "a data-collection issue cannot be ruled out, so treat the size of the drop as provisional while we keep monitoring."

The explanation that spread through SEO blogs is that Reddit deliberately blocked its entire domain from crawlers via robots.txt around 14 August. That is not true. I fetched `https://www.reddit.com/robots.txt` this morning and fetched the Internet Archive's snapshot from 14 April 2026. The two files are byte-identical, both ending in `User-agent: *` followed by `Disallow: /`. The blanket disallow was already four months old and names no exceptions at all, not even Google, which is why it never governed access in the first place: OpenAI reaches Reddit through a licensing agreement announced in May 2024 and still live, not through robots.txt permission.

Promptwatch itself, which produced the data, offers a different mechanism and shows its working. ChatGPT changed how it builds its own background searches on 8 August: use of the `site:` operator in fanout queries went from 0.37% to 16.8%, and average searches per response went from 1.08 to 1.83. Their sentence is "ChatGPT didn't stop searching the open web. It also started asking specific, named websites directly," and discussion platforms are rarely the target of a `site:` query about a business. This radar carried that fanout change on 21 August. The Reddit collapse is its downstream consequence, arriving three weeks later.

Method stated on the page: 1,146,483 questions, of which 531,889 had citations from all five engines (ChatGPT, Gemini, Perplexity, Google AI Overviews, Google AI Mode), collected 1 January to 30 June 2026 across 27 English-language markets, 84% US. Brand mentions were counted "whether or not it also links to that company." Headline figures, verbatim: "79.6% of the websites cited on a question appear on only one of the five AI engines"; domain-level agreement between engine pairs 10.9%; and the engines co-occur on "30.3% of the brands they name." Wellows' own limitation is that it measured overlap and not correlation, and collected each question once per engine, so run-to-run variance is not in it.

A number to refuse. Aggregators including authoritytech.io (1 September) attribute to Wellows a different set of figures: 596,723 prompts, 10.2% URL overlap, 17.9% domain overlap, 67.4% brand overlap. I could not find any of those four numbers on Wellows' own pages. Use the primary figures above or none.

ChatGPT.com is "up 48.38% year over year, now ranking ninth among all U.S. websites with 1.09 billion monthly visits." Bing is "down 50.43% year over year." Google remains first at 25.3 billion monthly visits and up 10.72%; Reddit is still third at 2.61 billion. Jarboe also reports direct traffic "up more than 157%" and above 78% of total traffic, but read the fine print: that figure is from one example publisher, not the panel, so it is an illustration and not a rate.