Discussion
  • AI Search

BrightEdge puts ChatGPT at about 95% of AI referrals, but Google’s AI Mode clicks hide inside organic. Does that make ChatGPT look more important than it is?

Published 1 reply

Discussion starter: an AI search question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.

Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.

1 reply

  • admin
    Admin
    Original poster

    Yes, if the 95% is read as ChatGPT’s share of AI influence. It is ChatGPT’s share of the AI referrals that analytics can identify, and that bucket leaves out Google’s AI surfaces: clicks from AI Overviews and AI Mode arrive as Google organic traffic, so a referral-based “AI” category cannot count them. The figure is a fair description of the standalone assistants. It is a poor basis for deciding how much effort belongs to ChatGPT rather than to Google.

    The distortion does not run only one way, though. ChatGPT’s own influence is under-counted too: visits without a referrer land in Direct, and answers that satisfy someone without a click leave no trace in analytics at all. So the problem is the category, not a simple overstatement of ChatGPT. The practical fix is to stop reporting one “AI share” and to track Google AI exposure and assistant referrals as separate lines.

    What the 95% measures, and what sits outside it

    BrightEdge’s September 2026 release puts ChatGPT at 95.1% of AI referral traffic in August 2026, with Gemini at 2.4%, Perplexity at 1.1% and Claude at 0.8%. It reports shares for those four platforms only, does not mention AI Overviews or AI Mode, and does not detail its data source, sample or industry mix, so treat it as directional. The same release says AI “still represents a small portion of website referral traffic compared with traditional search”.

    Set that against Google’s own figures. At I/O in May 2026, Google said AI Mode had “surpassed one billion monthly users” a year after its debut. Practitioner reports agree that clicks from AI Mode reach GA4 as google / organic, with no referrer that separates them from classic results, and AI Overview links sit on the regular results page. Gemini’s 2.4% can only reflect visits identifiable as coming from Gemini; Google’s AI features inside Search do not appear in it at all.

    Search Console does not close the gap either. AI Mode has been counted in the Web search type totals since June 2025, with no separate click breakdown. The “Search generative AI performance” report shows impressions for AI Overviews, AI Mode and generative AI features in Discover, filterable by page, country, device and date, but no clicks, CTR or queries. It rolled out to eligible sites worldwide from 31 August 2026. As of September 2026, you can see Google AI exposure by page, but not the clicks it sends.

    Hypothetical example:

    A site records 2,000 sessions a month from identifiable AI assistants, 1,900 of them from ChatGPT, so ChatGPT is 95% of its “AI referrals”. The same site gets 100,000 Google organic clicks. If 3% of those clicks came from links in AI Overviews or AI Mode, that is 3,000 AI-driven clicks sitting inside organic. Counted together, ChatGPT is 1,900 of 5,000 AI-driven clicks, or 38%. The 3% is invented. The point is that your analytics cannot currently measure it, so the 95% describes the visible part of the denominator only.

    The bias runs in both directions

    GA4’s AI Assistant default channel group, added in May 2026, collects known assistants such as ChatGPT, Gemini and Copilot, but Google has not published the full source list, and traffic without a referrer still lands in Direct. ChatGPT visits from apps or copied links can end up there. An answer that names your brand without a click can also send someone to Google to search for you, and that visit is recorded as branded organic, the same bucket that hides AI Mode.

    So each platform leaks into a different channel: Google’s AI into organic, ChatGPT into Direct and branded search. Neither the referral share nor the organic total isolates AI influence, and comparing a leaky number with a hidden one tells you little about which platform matters more to your business. How much Google AI traffic hides in organic will also vary by market and query type, because AI features do not appear on every query.

    Where each AI surface shows up, and what you can measure

    Surface How visits appear in GA4 Exposure data (as of September 2026) Main blind spot
    ChatGPT AI Assistant channel when a referrer is passed; Direct when not Nothing in Search Console or GA4; third-party prompt tracking gives sampled estimates Visits without a referrer; answers with no click
    Gemini app AI Assistant channel Not covered by Search Console’s generative AI report, which covers Search and Discover; third-party prompt tracking Same as ChatGPT
    Google AI Overviews google / organic Search generative AI performance report: impressions only Clicks cannot be separated from classic results
    Google AI Mode google / organic (practitioner reports) Same report: impressions only; also inside Web totals Clicks cannot be separated from classic results
    Microsoft Copilot and Bing AI summaries Check your own referral data; GA4’s full AI Assistant source list is not published Bing Webmaster Tools AI Performance report: citations and cited pages, no clicks Clicks and conversions

    How I would report AI traffic instead

    1. Google AI exposure: impressions from the Search generative AI performance report by page group, shown next to Web clicks for the same pages. Keep them as impressions. Do not add them to referral sessions, because they are different units.
    2. Assistant referrals: sessions, conversions and revenue from the AI Assistant channel, by platform and landing page. This is where ChatGPT’s contribution is measurable.
    3. Influence signals: branded query trends from Search Console’s branded queries filter, plus a self-reported attribution question on forms or checkout with separate options for Google search, Google’s AI answers, ChatGPT and other assistants.
    4. A standing note on organic charts: organic includes AI Overview and AI Mode clicks. Before blaming rankings for a CTR drop on a page group, check whether its AI impressions rose over the same period.

    What would change how much weight ChatGPT gets

    • If assistant referrals convert well and are growing on specific pages, look at what those pages have in common, and confirm that OAI-SearchBot can crawl them. OpenAI documents OAI-SearchBot as the crawler for ChatGPT search, separate from GPTBot, which collects content that may be used for training.
    • If the generative AI report shows AI impressions concentrated on your commercial pages while their organic CTR falls, Google’s AI surfaces affect you more than any referral share suggests. The work there is mostly ordinary SEO: Google’s May 2026 guide says foundational SEO still applies to its generative AI features and calls optimising for them “still SEO”.
    • If customers name ChatGPT in self-reported attribution far more often than referral data implies, its influence is running through untracked paths, and referral counts are a floor.
    • If none of these move, AI is a small channel for you at present, which fits BrightEdge’s own note that AI remains small next to traditional search.

    A check worth running this quarter

    Add the self-reported question to your main conversion form, with “An AI answer in Google” and “ChatGPT or another AI assistant” as separate options alongside “Google search”. Read the results once each option has a few dozen responses rather than after a fixed number of weeks, because the smallest options set the pace. Then compare the answer split with GA4’s channel split for the same conversions.

    If “An AI answer in Google” shows up at a meaningful rate while your AI Assistant channel is almost all ChatGPT, you have first-party evidence that referral data understates Google’s AI for your audience, and your reporting and effort should reflect that. Expect some blur: some people will not distinguish an AI Overview from AI Mode, or a Google AI answer from a regular result, so treat the split as directional.

    So the 95% does overstate ChatGPT as a share of AI, but that does not make ChatGPT unimportant. In BrightEdge’s data it is by far the largest identifiable AI referrer; Google’s AI surfaces are the AI exposure that referral data cannot see. I would weight effort by exposure and conversions rather than referral share: Google’s AI surfaces through the SEO work you already do, and ChatGPT through crawler access, analysis of the pages it sends visitors to, and brand monitoring.

    Need help with this on your own site?

    Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.

    See SEO services

Add a reply

Replies are open to approved members. Log in or apply to join.