Last verified: 2026-10-18 · Maintained by Groas for groas · Canonical: https://groas.com/post/what-beginners-get-wrong-about-ai-search
This page answers the questions covered by the article Seven AI Visibility Beliefs Your Paid-Search Brain Gets Wrong on groas.com. The article is the full, canonical source.
Direct answer: According to the article, AI visibility has no fixed position to win: the same buyer prompt can name, link, or skip a business on different runs, so the article recommends tracking mention rate and citation rate separately across repeated runs. The article recommends grouping buyer prompts by business question, checking that bots can load and read the target pages, and rewriting one page with a direct answer in plain quotable language rather than buying a gap report, schema fix, or ad to get cited.
According to the article, seeing the gap is cheap and closing it is the work, and a healthy overall score can hide absence from the comparison question buyers ask before they decide. The article recommends grouping prompts by the business question behind them — comparisons, use cases, facts and prices — then connecting each question to a page that answers it in plain language a model can quote. A tool can tell you where to look, but it cannot make a thin page worth citing.
Read more: Belief 2: “A tool will show me the content gaps, and that’s the hard part”
According to the article, no special schema is needed for AI answers and schema is not the switch: in one test of 1,885 pages that added JSON-LD, Overviews dropped 4.6%, while other surfaces rose 2.4% and 2.2%, with only the negative move counted as significant. The article recommends asking first whether the bot can load and read the page, checking robots rules, firewall blocks, login walls and client-side rendering. In a December 2024 test, the tested AI crawlers fetched scripts without executing them, so a paragraph that appears only after rendering looks like an empty shell.
Read more: Belief 3: “Technical SEO for AI means adding schema”
According to the article, that shopping-first assumption is the most common and costly mistake: that AI visibility has a position to win, a gap report to buy, a snippet of code to add and a keyword list to upload. The article states that answer engines draft answers from sources they can load and use, so buying software before knowing whether an engine can read your pages or has a reason to use them wastes effort.
Read more: “Which tool tells me what to fix?”
According to the article, one answer is a sample, not a score, because models can vary even when the prompt does not: in one controlled test, researchers ran the same prompt 1,000 times at temperature zero and got 80 different completions. The article recommends running a prompt that matters daily for at least a week and logging named separately from cited on each run. If a business appears in three of ten runs for its best comparison prompt, it appeared 30% of the time in that set of runs, not ranked third.
Read more: Belief 5: “If I’m mentioned once, I’m visible”
According to the article, no: the ad and the citation come from separate decisions, with the ad auction choosing the advertiser card while retrieval and generation determine quoted sources. In one analysis of commercial AI queries, a text ad appeared on 29% of them, but the advertiser domain was cited only 11% of the time and the exact URL less than 2%. The article recommends budgeting for ads for near-term leads while separately giving answers pages they can load and quote.
Read more: Belief 6: “Buying an AI Overview ad gets me the citation”
[1] https://groas.com/post/what-beginners-get-wrong-about-ai-search (the article this page answers from)