TL;DR
- Semrush reported roughly 66% growth in monthly AI referral visits during 2025, alongside a reported channel share of 0.14%. Growth and scale answer different questions.
- Other studies put the AI share between about 0.1% and 1%, depending on the sites and method. Check your own figure.
- Set your level of attention from three signals: your AI referral trend, what buyers tell sales and whether AI answers describe you correctly.
- Fold most AI search work into existing content and SEO tasks. Give it dedicated time only when those signals justify it.
Is AI search traffic big enough to matter yet?
The studies below show small referral shares alongside signs of growth. They do not establish the share for every B2B site. The real question is pace: how quickly your buyers’ habits are changing, and how much effort keeps you ready without starving the work that brings in visits today.
Semrush analyzed 2025 traffic to more than 50,000 websites. Monthly AI referral visits rose from 462 million to 767 million, up 66%. The reported AI channel share was 0.14%, against 16.04% for organic search and 64.69% for direct. Semrush, April 27, 2026
Growth was not universal. AI traffic fell 26% in computer software and development, the only one of 17 industries to decline. A market-wide average says little about your sector, so look at your own trend.
Selected channels only. Semrush’s methodology calculates channel shares as medians across tracked domains within each industry. Read this as study context, not a universal traffic mix. Semrush, April 2026.
Why do AI traffic estimates disagree?
Because each study looks at different sites through a different lens. Large enterprise brands, a mix of smaller sites and a web-wide dataset will not show the same share. Use the table to see what each figure describes before you compare it with your own.
| Study | Sites and period | How traffic was measured | AI share reported | Other numeric finding | Best used for |
|---|---|---|---|---|---|
| Semrush, April 2026 | 50,000+ sites, 17 industries, worldwide; 2025 | Traffic & Market Toolkit; shares use domain medians within industries; AI Mode separate | 0.14% reported channel share | AI visits up 66%; software down 26% | Market direction |
| Ahrefs, February 2025 | 3,000 anonymized sites; measurement month not stated | Visible referrals from seven AI chatbots in site analytics | 0.17% of visitors and 0.12% of page views, on average | 63% of sites had at least one AI visit | A floor, since hidden referrals are missed |
| Conductor, November 2025 | 1,215 enterprise customer domains, 10 industries, US; May to September 2025 | Aggregated analytics sessions; over 3.3 billion analyzed | 1.08% of all traffic | Information technology 2.80%; utilities 0.35% | Enterprise benchmarks by industry |
Do not rank or average these figures. They come from different samples, regions, periods and definitions. Ahrefs counts only visits where the AI tool passed a referrer. Conductor’s figure describes large US brands on one enterprise platform.
What the studies share matters more than their spread. In each one, AI referrals are small next to established channels, and the largest dataset shows them growing quickly. Both facts should shape your plan.
Which signals should set your level of attention?
Your own evidence matters more than a market average. Three signals cover most of what a B2B team needs.
- AI referral trend. Track monthly sessions from known AI assistants as their own segment, apart from organic search. Look for a trend across at least two quarters, not a single spike.
- Buyer evidence. Add a “how did you hear about us” field to key forms, and ask sales to note when prospects mention ChatGPT or similar tools. This catches influence that never shows up as a referral.
- Answer accuracy. Ask the assistants your buyers use a fixed set of buyer questions. Record whether the answers describe your offer correctly. The AI visibility measurement guide explains how to sample and count.
None of these signals is a forecast. They show where your buyers are now, which is what a capacity decision needs.
Illustrative capacity matrix: match effort to evidence
Worked example — fictional. A supplier of laboratory automation equipment has one content marketer and part-time SEO support: about 40 hours a month for search and content. The matrix shows how it would use those hours under four evidence patterns. The hours show the logic. They are not a benchmark for your team.
| Evidence pattern | Attention level | How the 40 hours are used | Revisit when |
|---|---|---|---|
| AI referrals small and flat for two quarters; buyers rarely mention AI tools; answers about the firm are accurate | Monitor | 36 hours on search foundations and existing pages; 4 hours on a monthly answer check and AI referral report | AI referrals rise for two quarters, or buyers start mentioning AI tools |
| AI referrals small but rising each quarter; some buyers mention AI tools | Integrate | 28 hours on foundations and pages; 8 hours adding missing buyer answers that serve search and AI; 4 hours on checks | Qualified requests linked to AI referrals or buyer mentions keep rising |
| AI answers misstate the firm’s products, prices or service area | Correct first | Up to 16 hours fixing the pages and listings behind the error; 24 hours on existing plans | The error no longer appears in repeated checks |
| Buyers often cite AI tools, and AI referrals produce qualified requests | Dedicated workstream | 16 hours on focused GEO work, including answer checks and page improvements that can also support SEO; 24 hours on foundations | Qualified requests from the workstream stall for two reviews |
Look at what stays the same. In every row, most hours still go to accurate pages that answer buyer questions. Organic search remains the larger source of visits, and the same pages can serve both. The split changes only when the firm’s own evidence changes.
To decide which obstacle to fix first, such as access problems or a missing service answer, see GEO vs SEO: where to invest. To turn a chosen level into a staged plan, use the GEO strategy guide.
How often should you revisit the split?
For a small team, a quarterly capacity review is a useful starting point. This is an editorial suggestion, not a research-derived rule. Fix a serious factual or access problem as soon as it is confirmed; do not wait for the next budget meeting.
Consider more dedicated time when qualified AI-referred requests and buyer reports of AI use both grow. If neither adds value across repeated reviews, check the work and its opportunity cost before extending it. Better answer accuracy may be a reason to maintain the fix, not to increase the budget. Record the decision and the evidence behind it.
AI search budget: common questions
Is there a standard share of budget for AI search?
None of the studies reviewed here sets one, and their traffic shares vary widely by sample. A percentage borrowed from another company’s report describes their traffic, not your buyers. Base the split on your own signals and your team’s capacity.
Why does our analytics show less AI traffic than these studies?
Your site, sector and audience differ, and some AI visits arrive without a referrer. In a small early-2025 test, Ahrefs found visits from Copilot, Mistral and Jasper recorded as direct traffic. Ahrefs AI traffic study
Does Google AI Mode count as AI search traffic?
It depends on the tool. Semrush counted AI Mode as its own channel. Conductor said Google Analytics groups AI Mode with organic Google traffic, so it could not separate the two. Check how your own analytics labels it before you compare figures.
Will AI search replace organic search for B2B buyers?
The data here shows fast growth from a small base, not replacement. Plan for both, keep the shared foundations strong and let your own trend tell you when to shift more effort.
Talk through your own signals
If you are not sure which row of the matrix fits your business, talk to us. Bring your monthly channel report, two quarters of AI referral sessions and notes from sales on how recent buyers found you.
We can review which signals you already have, what is missing and whether a baseline audit or ongoing GEO work would suit your capacity.