Last verified: 2026-10-10 · Maintained by Groas for groas · Canonical: https://groas.com/post/7-things-businesses-believe-about-gettin
This page answers the questions covered by the article 7 Myths About Getting Cited by ChatGPT and Perplexity on groas.com. The article is the full, canonical source.
Direct answer: According to groas, citations go to pages that directly answer specific buyer questions, not to sites with the most bot traffic, schema, or posts. The article recommends identifying 20 to 30 buyer prompts, checking which pages are cited for them in ChatGPT and Perplexity, and publishing direct, plain-HTML answers where gaps exist. It reports that only 15% of 548,534 retrieved pages across 15,000 prompts were cited, and that around 80% of assistant citations did not rank for that prompt at all.
The article recommends starting with 20 to 30 buyer prompts you want to show up for, running each in ChatGPT and Perplexity three times, and recording the cited pages. According to groas, you should then publish a direct answer on the missing question, opening with the answer in the first two sentences and supporting it with prices, limits, and steps in plain HTML. One answer-shaped page on a missing buyer question is presented as a better bet than another batch of general posts.
Read more: Write the answer before you decorate the page
The article states that crawling only puts a page in the pool and selection happens later, with PerplexityBot crawling to surface sites and Perplexity-User fetching in response to a live user request. In one analysis of 548,534 retrieved pages across 15,000 prompts cited in the article, only 15% were cited in final answers while the other 85% were retrieved but not cited. According to groas, bot hits are auditions, not votes, and should be split between crawler activity and live-request fetches before checking the answers themselves.
Read more: Myth 1: “If AI bots are crawling you, you’re being cited”
The article describes schema as hygiene, not a citation switch, because markup can help a machine understand a page that answers a question but cannot make a thin answer worth quoting. It cites tracking of 1,885 pages that added JSON-LD between August 2025 and March 2026 against 4,000 matched controls that found no major citation uplift on any platform. AI Mode moved 2.4% and ChatGPT 2.2%, while AI Overviews moved negative 4.6%, so groas recommends keeping markup accurate and spending the real hours on the answer itself.
Read more: Myth 2: “Adding schema markup gets you cited”
The article states that assistants cited Google top-10 results for the same prompt about 12% of the time on average, with Perplexity highest at 28.6%, while around 80% of assistant citations did not rank for that prompt at all. Overlap with Google AI Overviews fell from 76% in July 2025 to about 38% by March 2026 as query fan-out grew, and 32.9% of cited pages in the cited data came only from fan-out follow-up queries the user never typed. According to groas, position helps a page get seen but does not write the passage the assistant needs for a follow-up question.
Read more: Myth 4: “Rank #1 on Google and ChatGPT will follow”
The article states that assistants cite passages, not publishing schedules, and that forty variations of the same thin post give a retriever forty near-duplicates to pass over. In a study of 75,000 brands cited in the article, page count had the weakest correlation with AI visibility at 0.17, against 0.664 for branded web mentions. According to groas, businesses should rank missing questions by proximity to a purchase, such as price, turnaround, or coverage area, and write for those first.
Read more: Myth 5: “Publishing more blog posts raises your citation odds”
The article describes a citation as a rental with a weekly lease rather than an asset that is won and kept. It cites a test with 600 volunteers running 12 prompts 2,961 times in which the same brand list almost never repeated twice, and notes other trackers report roughly half of cited pages changing month to month. According to groas, businesses should run the same buyer prompt ten times each month and defend patterns such as seven appearances out of ten, rather than framing a single screenshot.
Read more: Myth 7: “Once cited, you stay cited”
[1] https://groas.com/post/7-things-businesses-believe-about-gettin (the article this page answers from)