Obsurfable

ChatGPT Handles One-Fifth of Google’s Search-Like Queries — But Sends 0.36% of Ahrefs’ Tracked Traffic

Obsurfable

Every quarterly board deck now asks the same question: how big is ChatGPT search compared to Google? The November 2026 answers depend on which yardstick you grab.

On November 3, 2026, Search Engine Journal’s Matt G. Southern reported comments from OpenAI consumer product lead Adam Fry: ChatGPT gets roughly one-fifth as many search-like queries as Google — counting each user question that triggers web search.

The same week, Ahrefs updated its public AI vs Search traffic benchmark (Nov. 3 refresh across 118,671 sites): Google sent 24.74% of tracked visits; ChatGPT sent 0.36%. AI assistants as a group: 0.46%.

Both statements can be true. Query volume measures how often people ask answer engines to search the web. Referral traffic measures how often those answers send someone to your site. AEO strategy needs both — and a third layer: mentions without clicks.

What Fry actually said

Fry spoke at a New York dinner hosted by OpenAI; journalist Alex Kantrowitz published the exchange via Big Technology on November 3.

OpenAI’s counting rule:

  • A search-like query is any ChatGPT question that invokes web search (Browse / search tool), not every chat message.
  • Under that definition, ChatGPT’s volume is about 20% of Google’s search-like volume (~1/5, sometimes quoted as “5× smaller”).

That is a behavioral comparison: how much retrieval ChatGPT runs versus Google Search.

It is not:

  • Total ChatGPT prompts (most are drafting, coding, casual chat).
  • Website referrals from ChatGPT.
  • Brand visibility (mentions in answers without links).

Earlier 2026 estimates used different denominators. Ahrefs’ February model inferred ~12% of Google search volume using Harvard/OpenAI prompt-mix research — a prompt classification method, not Fry’s web-search invocation rule. The new ~20% figure is OpenAI’s own definition, not a third-party crawl.

What Ahrefs’ October–November data shows

Ahrefs’ benchmark tracks sites using its free Web Analytics snippet — aggregated, not a census of the web.

October 2026 channel shares (1.205B visits in sample):

ChannelShare of visits
Direct31.08%
Search (all engines)27.55%
Unknown19.03%
Paid16.77%
Social7.78%
AI assistants0.46%
Email / referral~1.4% combined

Within search engines, Google rose to 24.74% (+0.78 pp MoM). Within AI assistants, ChatGPT holds 0.36% — largest single assistant source. Claude grew fastest (+75.9% visitors MoM) but from a tiny base; Gemini referrals fell 30.8% in the same window.

Interpretation for operators:

  • AI assistant traffic is growing in percentage terms (+0.03 pp on AI slice) but from a sub-one-percent absolute share.
  • Google organic remains the dominant discovery channel in this panel — even as AI Overviews expand (third parties estimated ~80% AI Overview presence by November 2026).

Marie Haynes’ November 2026 aggregation (cited in Search Engine Journal’s measurement roundup) sharpens the conversion story: ChatGPT sessions convert higher per session than Google organic in her 100-site GA4 panel — but Google still drove 96.5% of total conversions because volume dwarfs LLM referrals.

Why query share and traffic share diverge

Three mechanisms widen the gap between 20% query share and 0.36% traffic share:

1. Zero-click answers

ChatGPT often completes the task in the UI. Users never need a link. Fry’s metric counts queries that searched, not clicks issued. Google, too, loses clicks to AI Overviews — but classic results still ship ten blue links alongside many AI layouts.

2. Citation sparsity

Academic replay work in Chen and Koudas (arXiv 2610.15033) shows ChatGPT consults ~26 pages but cites ~3 per answer. Most consulted publishers receive no visit.

HubSpot’s November 3, 2026 guidance formalizes what Obsurfable has tracked: mentions (name in prose) and citations (linked sources) are different outcomes — only citations generate GA4 referrals. Fry’s query volume includes millions of answers where your brand may appear with no URL.

Planning implications by team

Brand / demand gen

  • Budget for share of voice, not only ChatGPT referral rows — use mention rate, citation rate, and competitor panels (how to measure AI visibility without fooling yourself).
  • Treat ~20% search-like query share as a top-of-funnel reach argument to leadership — even when traffic dashboards look flat.

SEO teams

  • Do not sunset SEO because AI converts well on thin traffic. Multiple November datasets (Haynes, SEO Works, Ahrefs) show organic still carries 20–40× the session volume of AI assistants in comparable GA4 panels.
  • Do add prompt libraries parallel to keyword lists — ChatGPT’s rewritten sub-queries rarely match SERPs (different indexes per engine).

Finance / CFO

  • Report three metrics in the same slide: (1) Fry-style query incidence in your category from panels, (2) referral sessions, (3) branded search lift after AI mentions — per LLM measurement in November 2026.

How this interacts with OpenAI’s distribution bets

Query share growth can outpace referral share if OpenAI optimizes for answer completeness:

Google’s response — more links in AI Overviews — tries to re-monetize attention for publishers. OpenAI’s Fry comparison says ChatGPT is already one-fifth the retrieval load of Google; it does not promise one-fifth the ad-supported click pool.

How Obsurfable fits

Referral analytics will undercount influence for years. Obsurfable records answers — including brands named without links — so teams can relate query growth to mention share even when Ahrefs shows 0.36% traffic. Match Explorer evidence to your company in the App; use Tasks to widen prompt coverage as ChatGPT search adoption rises.

FAQ

Did ChatGPT surpass 20% of all Google queries?

No. Fry compared search-like ChatGPT queries (web search invoked) to Google search volume — not every ChatGPT prompt, not every Google property.

Why do vendors quote Marie Haynes’ 4.5× conversion multiplier?

Session-level conversion on tiny ChatGPT traffic can look stellar. Haynes herself emphasizes Google still drives most conversions in absolute terms.

Should we reallocate SEO budget to ChatGPT because of the 1/5 statistic?

Reallocate measurement and content toward answer engines — not necessarily paid media mix until category panels show citation/mention lift tied to revenue proxies.

Is Gemini’s referral decline in Ahrefs meaningful?

It reflects October 2026 panel dynamics, not necessarily global share. Track your GA4 and prompt panels.

Bottom line

OpenAI’s ~one-fifth search-like query ratio says ChatGPT is a major retrieval channel by usage. Ahrefs’ 0.36% referral share says your site still lives on Google clicks. AEO programs that only watch Analytics will look like failures while competitors accumulate mentions. Measure queries, mentions, citations, and branded search together — and tell leadership which number answers which question.