Marketing AI

Is AI Search Traffic Worth Chasing? Clicks Lost vs Leads Won

AI answers are linked to fewer clicks from search (one randomised study found a cut), and the visits AI assistants send are few but convert better. Whether that trade pays depends on your own numbers, which most teams are not yet tracking.

Fine-line drawing of a magnifying lens over a wide field of dots, a few dots flowing through a narrow channel to a house with an amber door.

TL;DRThe 8-minute read in 60 seconds

AI search traffic is worth chasing for its quality, not its volume. People click through about half as often when an AI summary shows, but AI visitors tend to buy or sign up more.

  1. In Pew's data, people clicked a result in 8% of visits with an AI summary, and in 15% without one.
  2. On publisher sites Microsoft Clarity studied, AI visitors signed up at 1.66% against 0.15% from search. That is an upper bound from newsletter sign-ups, not a B2B lead benchmark.
  3. Whether AI visits make up for lost clicks depends on how well they turn into leads for you. The calculator shows your break-even.
  4. Start by tracking it: the GA4 AI Assistant channel, Search Console's AI report and Bing's AI Performance report.

AI search traffic is worth chasing for its quality, not its volume. People click a result about half as often when an AI summary appears, while the few visits AI assistants send convert better than search visits.

On cautious assumptions the second effect does not yet pay for the first. On the most favourable published figure it does. This guide sizes both with sourced numbers, shows how to measure your own, and helps marketing and sales leads decide where the next dollar should go.

At a glance: clicks lost, leads won

Three numbers and one first step, each sourced below.

CLICKS LOSTAbout halfPeople click a result in 8% ofvisits with an AI summary, 15%without (Pew).LEADS WONFew, but betterAI tools sent under 1% ofvisits. Sign-ups ran 1.66%against 0.15% from search(Clarity, publisher sites).DO FIRSTMeasure itCheck GA4, Search Console,Bing and a CRM lead-sourcefield before moving budget.Illustrative 10,000-visit site: about 6 fewer leads a month on cautious rates, about 2.5 more onthe most favourable published gap.
CLICKS LOSTAbout halfPeople click a result in 8% of visitswith an AI summary, 15% without (Pew).LEADS WONFew, but betterAI tools sent under 1% of visits.Sign-ups ran 1.66% against 0.15% fromsearch (Clarity, publisher sites).DO FIRSTMeasure itCheck GA4, Search Console, Bing and aCRM lead-source field before movingbudget.Illustrative 10,000-visit site: about 6fewer leads a month on cautious rates,about 2.5 more on the most favourablepublished gap.
Sources: Pew Research Center (22 July 2025); Microsoft Clarity (6 November 2025, publisher and news sites); the illustrative example is worked out in the calculator section.

The scale is clear. Google says AI Overviews has "over 2.5 billion monthly active users" (Google I/O keynote, 19 May 2026) and that AI Mode has passed 1 billion (first said at I/O on 19 May 2026 and repeated in the Alphabet Q2 remarks, 22 July 2026).

The 2026 AI-search dashboard

Eight numbers that frame the decision, each with its source and date.

2.5B+AI Overviews monthlyusersGoogle, May 20261B+AI Mode monthly usersGoogle, July 202618%Searches that showedan AI summaryPew, July 20258% vs 15%Visits with a click,AI summary vs nonePew, July 20251%Visits with a clickinside the AI summaryPew, July 2025−18.8 ptsChange inclick-through underforced AI ModePreprint, Aug 2026under 1%Share of visits fromAI assistantsClarity, Nov 20251.66% vs 0.15%Sign-up rate, AIreferrals vs searchClarity, Nov 2025
2.5B+AI Overviewsmonthly usersGoogle, May 20261B+AI Mode monthlyusersGoogle, July 202618%Searches thatshowed an AIsummaryPew, July 20258% vs 15%Visits with aclick, AI summaryvs nonePew, July 20251%Visits with aclick inside theAI summaryPew, July 2025−18.8 ptsChange inclick-throughunder forced AIModePreprint, Aug 2026under 1%Share of visitsfrom AIassistantsClarity, Nov 20251.66% vs 0.15%Sign-up rate, AIreferrals vssearchClarity, Nov 2025
Sources: Google I/O keynote (19 May 2026) and Q2 remarks (22 July 2026); Pew Research Center (22 July 2025); Wang et al., arXiv preprint (18 August 2026); Microsoft Clarity (6 November 2025).

How many clicks do AI answers take?

When an AI summary appears, people click through to websites about half as often, and the strongest evidence is independent. Pew Research Center tracked 900 US adults through 68,879 Google searches in March 2025: users clicked a traditional result in 8% of visits with an AI summary and 15% without, and clicked a link inside the summary in just 1%.

Interactive chart

Half as many visits end in a click when an AI summary shows

Share of visits to a Google results page, March 2025, with and without an AI summary. Users clicked a link inside the summary in 1% of visits.

With an AI summary, users clicked a traditional result in 8% of visits against 15% without one, and ended their browsing session on 26% of such pages against 16%.

Share of visits, with vs without an AI summary (Pew, March 2025)
OutcomeWith an AI summaryWithout one
Clicked a traditional result8%15%
Ended the browsing session26%16%
Source: Pew Research Center, 22 July 2025 (900 US adults, 68,879 searches; independent survey panel, not peer-reviewed).

A newer field experiment points the same way. Researchers at the University of Pennsylvania and Northeastern University randomly assigned US Chrome users (1,100 in the analysis) to different Google layouts in a ten-day study run in March 2026. Compared with standard Google Search, forcing AI Mode cut click-through to other sites by 18.8 percentage points (arXiv preprint, 18 August 2026). It is a preprint, not yet peer-reviewed, but it is a randomised test, which is stronger evidence than a log of what people happened to click.

Does AI traffic convert better?

Yes, in the largest public dataset, but from a small base. AI assistants sent under 1% of visits to the publisher sites Microsoft studied, and those visitors signed up about 11 times as often as search visitors. Microsoft Clarity's study (6 November 2025) covered 1,277 publisher and news domains. The subscription gap was smaller, 1.34% against 0.55% (about 2.4 times), and Clarity's own headline across its measures says "3x".

Interactive chart

AI referrals convert better, from a small base

Conversion by channel on publisher and news sites, one month. B2B sites will differ. Toggle a series to compare.

AI-assistant visitors signed up at 1.66% against 0.15% from search, and subscribed at 1.34% against 0.55%. Direct traffic converted at 0.13% and 0.41%, social at 0.46% and 0.37%.

Sign-up and subscription rate by channel (Microsoft Clarity, publisher sites)
ChannelSign-up rateSubscription rate
AI assistants1.66%1.34%
Search0.15%0.55%
Direct0.13%0.41%
Social0.46%0.37%
Source: Microsoft Clarity, 6 November 2025 (1,277 publisher and news domains; one month of smart-event data). AI assistants were under 1% of visits.

Two cautions before you plan around this. These are publisher sites, where a sign-up is a newsletter, not a sales lead. And AI visitors may arrive later in their research than a search visitor.

Google, for its part, says total organic click volume from Search to websites "has been relatively stable year-over-year" and that it sends "slightly more quality clicks to websites than a year ago". Google does not define "quality" on that page, and we read it as a claim about click value, not a measured figure (Google, 6 August 2025). Google has published no numbers behind either claim.

Do the leads won make up for the clicks lost?

On cautious assumptions, not yet; on the most favourable published figure, yes. The answer turns on how much better AI visitors convert for you. The calculator below makes that trade explicit with your own numbers, starting from Pew's click gap and illustrative B2B rates.

Illustrative example: a B2B site gets 10,000 organic search visits a month. If 18% of those visits come from queries that now show an AI answer (Pew's share of searches with a summary) and those lose 47% of their clicks (Pew's 8% against 15%), the site loses about 846 visits. At a 1% lead rate that is about 8.5 leads.

Cautious case: a hundred AI-assistant visits converting at 2.4% (Clarity's 2.4-fold subscription gap applied to that 1%) add 2.4 leads. Net: about 6 fewer leads a month, and break-even needs about 353 AI visits, three and a half times the default.

Favourable case (an upper bound): Clarity's data come from publisher and news sign-ups, usually newsletters, so they are not a B2B lead benchmark and this case is illustrative only. Use its 11-fold sign-up gap instead (an 11% AI lead rate) and the same 100 visits bring 11 leads, about 2.5 more than were lost. Which gap holds for your leads is the number to measure.

Calculator

Clicks lost vs leads won, on your numbers

Visits lost = organic visits x share on queries with an AI answer x click loss. Leads lost = visits lost x organic lead rate. Leads won = AI-assistant visits x AI lead rate. Break-even = leads lost / AI lead rate.

10,000
18%
47%
100
1%
2.4%
500
Visits lost per month846
Leads lost8.5
Leads won from AI2.4
Net leads per month−6.1
Net lead value per month−3,030
AI visits needed to break even353

Defaults come from older data: Pew (March 2025) and Clarity (November 2025). Replace them with your own numbers. Sourced defaults: 18% (Pew, share of searches with an AI summary in March 2025, applied here to visits as an assumption) and 47% (Pew, 8% against 15% of visits with a click). Illustrative defaults, not measured: 100 AI visits (1% of organic visits; Clarity found AI under 1% of all traffic), 1% organic lead rate, 2.4% AI lead rate (Clarity's subscription gap; its sign-up gap would give 11%), a value of 500 per lead (unitless). Use your own GA4 and CRM numbers. Nothing you enter leaves your browser.

How do you measure AI search on your own site?

Use three free reports and one CRM field: Google Search Console's generative AI report, Bing's AI Performance report, the GA4 AI Assistant channel and a lead-source option for AI. Each one sees a different slice, and none sees everything. The diagram shows where each slice sits on the path from an AI answer to a lead.

From an AI answer to a lead, and where you can see each step

Two paths. Only the last two steps of the assistant path show up in your own analytics today.

GOOGLE AI OVERVIEW OR AI MODE1. AnswerAI Overview or AI Modeshows an answer2. ClickClicks fall: 8% ofvisits with a summary,15% without3. VisitArrives as Googleorganic search4. LeadYour organic lead rateAI ASSISTANT (CHATGPT, GEMINI, COPILOT)1. AnswerAn assistant answers andmay cite you2. ClickFew follow the link:under 1% of visits3. VisitArrives in the GA4 AIAssistant channel4. LeadSign-ups ran 1.66%against 0.15% onpublisher sitesWHERE YOU CAN SEE IT1. AnswerSearch Console AI reportand Bing AI Performance:impressions andcitations, no clicks2. ClickNo separate report: AIOverview clicks sitinside the overallSearch Console report3. VisitGA4: assistant visitsonly, not AI Overviewsor AI Mode4. LeadCRM lead-source field:you must add it
1. AnswerGOOGLEAI Overview or AI Mode shows an answerASSISTANTAn assistant answers and may cite youSEEN INSearch Console AI report and Bing AIPerformance: impressions and citations,no clicks2. ClickGOOGLEClicks fall: 8% of visits with asummary, 15% withoutASSISTANTFew follow the link: under 1% of visitsSEEN INNo separate report: AI Overview clickssit inside the overall Search Consolereport3. VisitGOOGLEArrives as Google organic searchASSISTANTArrives in the GA4 AI Assistant channelSEEN INGA4: assistant visits only, not AIOverviews or AI Mode4. LeadGOOGLEYour organic lead rateASSISTANTSign-ups ran 1.66% against 0.15% onpublisher sitesSEEN INCRM lead-source field: you must add it
Sources: Pew Research Center (22 July 2025); Microsoft Clarity (6 November 2025); Google Search Central (June 2026); Bing Webmaster blog (June 2026); Google Analytics help (GA4 default channel group).

The table shows what each source can and cannot tell you as of 4 October 2026.

Report Shows Does not show
Search Console, generative AI report AI Overviews and AI Mode impressions by page, country and device Clicks or queries
Bing Webmaster Tools, AI Performance (launched as a public preview in February 2026) Citations, average cited pages, a sample of grounding queries, citation share Traffic or clicks (not part of the report)
GA4, AI Assistant channel Visits from assistants such as ChatGPT, Gemini, DeepSeek, Copilot and Grok Visits from AI Overviews or AI Mode
Your CRM Which leads and deals started in an AI answer Nothing, until you add the field

Sources: Google Search Central, June 2026, Bing Webmaster blog, June 2026 and GA4 default channel group help.

One trap is worth avoiding. A loose pattern for an AI channel, such as any source containing "ai", also catches sources such as gmail.com and mail.google.com in a plain regex test. Origin Pi's suggested rule below matches only the assistant domains, including subdomains; GA4 sources are lowercase, as the rule expects.

Copy this rule

A custom GA4 channel for AI assistants

Create it in a custom channel group, set the condition to Session source matches regex, and place it above Referral. Tested on 4 October 2026 against sample sources.

Channel name: AI assistants

Condition: Session source matches regex

^(.+\.)?(chatgpt\.com|openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai|chat\.deepseek\.com|grok\.com|meta\.ai)$

Check it before you trust it

  • Matches: chatgpt.com, chat.openai.com, perplexity.ai with or without www, gemini.google.com, claude.ai.
  • Does not match: gmail.com, mail.google.com, notchatgpt.com.
  • Add any new assistant domain you see under Referral.

AI Overviews and AI Mode visits arrive as Google organic search, so no source rule can separate them.

Checklist

Can you see your AI search traffic?

Tick what is already set up. Nothing you enter leaves your browser.

Analytics
Search consoles
Sales
  • 0 to 39%: Flying blind. Turn on the free reports first.
  • 40 to 74%: Partly measured. Close the gaps before you move budget.
  • 75 to 100%: Measured. Size the trade with your own numbers monthly.

Weighted: the GA4 channel and the CRM field count most, because they connect visits to leads.

What changed in 2026, and what does the court ruling mean for you?

In 2026, Google and OpenAI turned AI answers into surfaces with ads and reports, and a US court declined to make Google send publishers traffic. On 30 September 2026, Judge Amit Mehta dismissed the antitrust suits that Chegg and Penske Media brought over AI Overviews. On their claim that Google owed them traffic in return for free content, he wrote:

But an expectation is not an agreement. It is simply how a general search engine works.

Judge Amit Mehta, Memorandum Opinion, US District Court for the District of Columbia, 30 September 2026, p. 20. The same opinion decides the Penske Media case.

The order is appealable. This is general information, not legal advice.

Timeline

How answers became ad and report surfaces

Select an event to see what it changes for your traffic.

  1. 22 Jul 2025

    Pew measures the click gap

    8% of visits with a click when an AI summary shows, 15% without.

  2. 6 Aug 2025

    Google claims quality clicks

    Google says AI search sends slightly more quality clicks, without publishing numbers.

  3. 6 Nov 2025

    Clarity: AI visits convert

    Sign-ups at 1.66% from AI assistants against 0.15% from search, on publisher sites.

  4. 13 May 2026

    GA4 adds an AI Assistant channel

    Visits from ChatGPT, Gemini and other assistants get their own default channel.

  5. 19 May 2026

    AI Overviews passes 2.5 billion users

    Google also says AI Mode has passed 1 billion monthly users (repeated in its Q2 remarks on 22 July 2026).

  6. 20 May 2026

    Ads tested in AI Mode

    Google announces it is testing new ad types in AI Mode, including Highlighted Answers in recommendation lists.

  7. 18 Aug 2026

    Randomised test of AI Mode

    A preprint finds forced AI Mode cut click-through by 18.8 percentage points.

  8. 31 Aug 2026

    AI report and opt-out for all sites

    Search Console's generative AI report and the AI-features opt-out reach every site.

  9. 23 Sep 2026

    ChatGPT ads reach Southeast Asia

    OpenAI says it is expanding ChatGPT ads to Southeast Asia and Taiwan, bringing the total to more than 60 countries.

  10. 30 Sep 2026

    Court dismisses publisher suits

    Judge Mehta dismisses the Chegg and Penske Media cases over AI Overviews; the order is appealable.

Sources: Pew; Google blog, Search Central and GA4 help; Microsoft Clarity; arXiv 2608.18352 (preprint); OpenAI; CourtListener.

What does that mean in practice? Two things. First, you now have a lever Google built for site owners: Google added a Search Console toggle that lets a site opt out of its generative AI features (AI Overviews and AI Mode), and says "Sites that opt out will not receive traffic or impressions from our generative AI features" (Google, updated 31 August 2026). For most businesses that trades away visibility where buyers now research, so it rarely pays. Second, the answer surfaces now carry ads. Google is testing ads in AI Mode, including Highlighted Answers in recommendation lists, and OpenAI said on 23 September 2026 that ChatGPT ads were expanding to Southeast Asia and Taiwan. Paid is now a way to be present inside the answer, not only beside it.

Where should your next marketing spend go?

Fix tracking first, then put money where your buyers already ask. Earn citations if AI answers cover your category, test ads inside AI answers if your market has them, and defend paid search where its clicks still convert. The decision tool walks through that order.

Decision tool

Your next move on AI search

Up to four questions. A guide, not an audit. Nothing you choose leaves your browser.

  1. If you cannot yet see AI-assistant visits and AI impressions, fix tracking first.
  2. If AI answers cover your buyers' questions and your pages are rarely cited, earn citations.
  3. If you are already cited and AI ads run in your market, run a small, tagged AI ads test against paid search.
  4. If AI answers rarely cover your buyers' questions, or no AI ads run in your market, defend paid and organic search and revisit monthly.

The roles split cleanly. The marketing lead owns the trade: which pages to make citable and whether to test ads inside answers. The analytics owner owns the three reports and the custom channel. Sales owns the lead-source field, because an AI-referred lead that is logged as "direct" disappears from the case. Our post on measuring AI answer citations covers the citation side in detail.

Where Origin Pi stands

We think AI search is worth chasing as a source of better-qualified leads, not as a replacement for search volume, and only once you can measure it. The public data says the clicks are shrinking and the AI visits that remain convert better, but publisher data is not your data.

So we start with measurement: the GA4 channel, the console reports and a lead-source field that sales actually fills in. Then we make sure an AI-referred lead does not stall. We design and build agent-ready intake that answers AI-referred enquiries quickly, routes them to the right person with a confirm step before any commitment, and keeps an audit trail of where each lead came from. Our marketing AI work connects that intake to your content and campaigns. For citation-first content and ads inside AI answers, the card below links to Origin Pi's marketing consultancy brand leapbuzz.

Your first step: open GA4 today, find the AI Assistant channel, and compare its lead rate with organic search over the last 90 days. That one comparison tells you whether the trade is worth chasing for you.

Next step

Turn AI answers into leads you can measure

leapbuzz

AI visibility and paid tests

Earn citations in AI answers and test ads inside them with Origin Pi's marketing consultancy brand leapbuzz.

See leapbuzz visibility optimization

Origin Pi

AI-lead intake and routing

We build intake that tags AI-referred leads, answers them fast and routes them to the right person with an audit trail.

Ask for an intake review

Frequently asked questions

How much traffic do AI Overviews take?
Independent data says users click through about half as often when an AI summary appears. Pew Research Center's March 2025 browsing data from 900 US adults found users clicked a traditional result in 8% of visits with an AI summary and 15% without. A randomised University of Pennsylvania and Northeastern field experiment (arXiv preprint, August 2026) found that forcing Google AI Mode cut click-through to other sites by 18.8 percentage points.
Does ChatGPT traffic convert better than Google traffic?
In the largest public dataset, yes, from a small base. Microsoft Clarity's study of 1,277 publisher and news domains (6 November 2025) found AI-assistant visitors signed up at 1.66% against 0.15% from search, and subscribed at 1.34% against 0.55%. AI assistants were under 1% of visits. B2B sites will differ, so measure your own.
How do I track AI referral traffic in Google Analytics 4?
Check the AI Assistant default channel that GA4 added on 13 May 2026. Google says it covers sources like ChatGPT, Gemini, DeepSeek, Copilot and Grok and excludes AI Overviews and AI Mode. Add a custom channel group above Referral for any assistant it misses, and add an AI option to your CRM's lead-source field.
Can I see AI Overview clicks in Google Search Console?
Not separately. Search Console's generative AI report, which Google says reached all websites by 31 August 2026, shows impressions in AI Overviews and AI Mode by page, country and device, but not clicks or queries. Those clicks stay inside the overall performance report, so pair it with GA4 and your CRM.
Can publishers force Google to send them traffic?
Not under US antitrust law, on the first ruling. On 30 September 2026, Judge Amit Mehta of the US District Court for the District of Columbia dismissed the Chegg and Penske Media suits over AI Overviews, writing that an expectation of traffic is not an agreement. The order is appealable, so the question may return.

Next step

Ready to make your business agent-ready?

20 minutes on your sector, your systems, and where this applies. No deck, no templates.

Talk to us →