A channel with few visits may bring valuable enquiries, or it may not. Do not base decisions solely on traffic share or another company's result.
In this article I pull together public data available in 2026 and explain why AI referral traffic deserves separate measurement. This is a decision and channel-value article; the implementation steps live in a separate guide. No magic, no fear-selling.
In June 2025, Ahrefs published its own website data for the preceding 30 days: AI accounted for 0.5% of total traffic and 12.1% of all signups. This does not mean 12.1% of AI visitors signed up, nor is it a Czech or current market average. The case motivates measurement, not copying its ratio into forecasts.
How much traffic AI search actually sends
The share varies by website, audience and period. Separate referral source, signup count and revenue. The following example preserves historical Ahrefs figures rather than estimating the current market.
Worked example: Small traffic share, larger signup share
SOURCE DATA · AHREFS.COM
| Metric | AI share |
|---|---|
| Website traffic | 0.5% |
| Website signups | 12.1% |
Different denominators. Neither number is a conversion rate or market average.
What the Ahrefs case tells us
The original Ahrefs article of 16 June 2025 concerns its own website and a limited period. Traffic share and signup share have different denominators. The result does not prove that referral source caused stronger interest.
One possible explanation is stronger purchase intent: a user may ask follow-up questions and compare options in the AI interface, then visit a website later in the decision process. That is not a rule. Visit value varies by industry, query, offer and measurement setup.
AI answers also tend to use a limited set of mentions and sources. Their composition changes by platform, prompt and time, so a one-off screenshot is not reliable visibility measurement.
"But AI isn't sending me anyone"
Several explanations are possible: the channel sends no visits, measurement misses them, or visits are misclassified. Check measurement first.
First: you are not measuring it separately. Some referrals can be grouped in GA4 by source domain (chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai); some visits may still lack a reliable referrer. Treat what you can observe as a minimum, not a complete total.
Second: there are not enough discoverable inputs. Accurate facts, accessible content, consistent business information and trustworthy mentions give search and AI systems better source material. They do not guarantee a citation. The implementation is covered in the separate guide to getting your business into AI search.
What these numbers mean for a business decision
What the numbers mean for a decision
- Separate AI referrals track AI visits in analytics on their own
- Do not import a benchmark someone else's numbers are not your data
- Build the foundation the same one as for good SEO
Compare visits, inquiries and conversion rate with organic, direct and paid channels. Your offer, audience and measurement differ, so decisions also need case counts and enquiry quality. Discoverable pages, accurate facts, first-hand experience and trustworthy sources help people and machines, but they do not guarantee a citation or ranking.
The detailed technical, content, entity and measurement checklist is in the separate guide to getting your business into AI search. This keeps this page focused on data and decisions rather than duplicating the implementation guide.
Search Console describes performance in Google Search, not every AI recommendation. Record supported citations, identifiable visits and enquiries separately. SEO services do not guarantee citations or rankings.
Worked example: Business goal before channel
ORIGINAL EXAMPLE · FICTIONAL DATA
| Input | Example state | Decision |
|---|---|---|
| Goal | More suitable enquiries | Measure qualification, not just visits |
| Capacity | Room for a new project | Check timing with sales |
| AI visits | Referrer only sometimes identifiable | Unknown is not zero |
| Investment | Limited test budget | Evaluate own data |
Someone else's benchmark is not your revenue forecast.
When it tips
There is no need to wait for AI search to become "mainstream", or to invest blindly. Keep the website technically and editorially discoverable, establish measurement and use your own data to decide how much priority the channel deserves. There is no general guarantee for stable results or time to a first citation.
Summary
Evaluate AI visits using your own data, its completeness and commercial quality. The historical Ahrefs case illustrates a possible difference between volume and value; it does not prove universally higher conversion.
Sources and links
-
Ahrefs: own website, 30 days preceding publication on 16 June 2025
-
GEO: Generative Engine Optimization (Princeton, KDD '24), peer-reviewed study of citability techniques