ChatGPT does not search the web for every answer—and the gap between being mentioned and being retrieved is becoming a core marketing challenge. In a new analysis of 66.6 million free, logged-out GPT-5.6 Luna responses, Ahrefs found that just 14.4% triggered a web search during August and September 2026. When ChatGPT did search, it ran about 1.6 related queries on average. The findings suggest brands need two kinds of visibility: a reputation strong enough to be recognized and authoritative information ready when an AI system looks for evidence. Ahrefs’ study is a useful signal, not a universal forecast: it covers one model and a specific set of logged-out sessions.

AI visibility now has two different routes

For marketers, the headline is not simply “ChatGPT searches less.” The study shows a fork in the answer process: many sampled responses were produced without a live web search, while a smaller share triggered retrieval and follow-up queries. The data does not establish how often a brand appeared in answers that did not browse, nor does it prove what caused a brand to be remembered. It does show why a citation-only scorecard is incomplete: a company can be absent from a live-search citation and still be named—or omitted—in an answer.

Scope matters. Ahrefs examined free, logged-out GPT-5.6 Luna sessions across August and September, not every ChatGPT user, paid tier, model, or query type. Its citation trends came from a separate U.S. dataset. Search Engine Journal’s coverage notes that other studies report different retrieval rates and that the setups are not directly comparable. Treat 14.4% as a snapshot that should inform testing, not as a fixed property of ChatGPT.

Original evidence is gaining ground over summaries

The citation data points to a second shift: AI answers may be leaning more toward sources that originate information. In Ahrefs’ separate U.S. tracking, Reddit appeared in about one in five ChatGPT answers in March and one in 43 by late September; Wikipedia moved from roughly one in seven to one in 35. Research organizations, regulators, brand sites, support pages, and hands-on testers gained ground, while some explainers and aggregators declined. Ahrefs describes the pattern as a move from sites that summarize information toward the sources that produce it.

That does not make third-party coverage irrelevant. For factual questions about pricing, features, or company details, 84.4% of prompts in Ahrefs’ 20,000-prompt U.S. test cited at least one official source. For comparison questions, that fell to 68.3%, and official websites made up a smaller share of citations. The practical distinction is clear: keep your own facts complete and verifiable, while earning credible outside reviews and comparisons that help buyers weigh alternatives.

Make your brand the clearest source of its own facts

Start with the questions customers ask before they contact sales: current prices or pricing rules, product specifications, availability, service terms, compatibility, returns, and support. Publish the answers in accessible, plain text on stable pages. Keep those facts consistent across your site and trusted business profiles; label updates and link claims to primary evidence where relevant. A polished campaign page cannot compensate for conflicting or missing details when an AI system seeks a direct answer.

Then add something competitors cannot simply restate. Publish original tests, benchmark data, customer outcomes with methodology, expert commentary, or documented product changes. For comparison-driven searches, make differences concrete and fair: identify who each option suits, disclose limitations, and cite supporting evidence. Do not turn every page into a sales pitch. Independent reviews and industry sources can provide the context a brand’s own site cannot.

A practical 30-day AI search plan

  • Week 1: Build a test set of 20–30 prompts across factual, comparison, pricing, review, and “best for” intent. Record whether ChatGPT searches, mentions the brand, cites the site, and names alternatives.
  • Week 2: Audit the pages behind those answers. Fix stale claims, missing specifications, ambiguous pricing, hard-to-find support details, and contradictions across pages.
  • Week 3: Create or improve one original proof asset and one transparent comparison page. Ask customers and subject-matter experts to validate factual claims before publication.
  • Week 4: Repeat the prompt test, compare citations and brand mentions separately, and track qualified visits, leads, and branded demand. Keep the model, account type, date, and prompt constant so the comparison is meaningful.

AI search is not a single ranking contest. Brands need both recognizable authority and dependable first-party evidence, then measurement that separates mentions, citations, and business outcomes. Real Internet Sales helps businesses build that strategy. Call 803-708-5514 or visit realinternetsales.com.