Key Takeaways
- Organic traffic is down, with a recent Brainlabs study showing that 85% of its client sites saw organic sessions fall between 2025 and 2026.
- Though a smaller segment, AI referrals are 1.5x more likely to take actions like purchases and signups than organic search visitors.
- AI platforms are sending some visitors while also crawling and pulling content from websites, making it harder to understand website traffic from visitor numbers alone.
Everyone knows that AI is changing the way people search the web. Many of us have done our fair share of asking a chatbot for recommendations, explanations, comparisons, or just a quick answer instead of opening a dozen search results.
But relying on this may be hurting the raw numbers business owners use to measure whether anyone’s visiting their sites, suggests new data from Brainlabs, published by Digiday. The findings show how much AI is changing search traffic for its clients: 46 of 54 sites lost organic traffic between January 2025 and April 2026, with sessions falling 10.5%.
Organic Search Traffic Is Falling
Traffic fell from 140.1M to 125.4M between 2025-2026
There is a silver lining, though: Those who came from ChatGPT, Copilot, Gemini, and Perplexity all purchased, signed up for a newsletter, scrolled the whole page, or other typical buyer-journey actions at 1.5x the rate of visitors who came through traditional, organic search.
This means that while websites may be getting fewer visitors from search, they may also be getting better visitors.
Brainlabs analysts theorize it’s because AI referrals have stronger intent, which helps offset the fact that they’re still a relatively small source of traffic compared to traditional search. For reference, the study showed that AI referrals averaged ~200,000 sessions per month, compared with ~20 million monthly sessions from organic traffic.
It makes sense. A person who Googles something general gets the answer directly through AI Overviews, maybe only needing a quick comparison, single fact, or a simple definition. But someone who’s coming from ChatGPT may have already spent several prompts researching something, asking questions, narrowing down choices.
So by the time that person reaches a website, they may be a lot more willing to do their part in the buyer’s journey.
There’s More to AI Traffic Than Referrals
Referral traffic is just one type of traffic: While AI platforms are sending visitors to websites, they’re also crawling and extracting their content, often without sending anything back.
Microsoft Clarity recently added an AI Scrape-to-Referral Ratio to its Bot Analytics dashboard, which compares AI scraping activity with referral traffic to see how much crawling is actually translating into visitors.
None of this is to say that a website can’t benefit from being included in AI answers, even if those answers don’t lead to a click or conversion. A mention or citation can still introduce someone to a business, influence a later search, or contribute to brand recognition, and these are things that aren’t measured by even the best analytical tools. In fact, a recent study by Scrunch found that after an AI platform recommended a brand to someone who wasn’t already using it, that person was more likely within a week to search for the brand, visit its website, and view its products.
Still, AI has complicated how we should now be defining traffic value. A click and crawler is measurable, but the influence of an AI recommendation that leads someone to search for a brand days later? Not so much.
What This Means for Hosting Providers
Most hosting providers offer some type of analytics or SEO dashboard, meaning clients may notice when their monthly visitor counts decline and assume something is wrong with their website. But a drop in human visitors doesn’t necessarily mean a site requires fewer hosting resources.
An eCommerce site, for example, may receive fewer but more “qualified” visitors while still relying on fast databases, caching, and reliable infrastructure, to support any transactions.
At the same time, AI crawlers can create a decent amount of traffic that never appears in traditional website analytics. Depending on how a site is configured, crawlers can consume CPU, trigger database queries, and repeatedly hit the server instead of being handled by a CDN or cache layer. A website’s visible traffic may be declining even while the infrastructure is working just as hard or even harder.
Depending on a customer’s hosting setup and billing model, increased bot traffic could easily contribute to higher bandwidth usage, greater compute consumption, or the need to upgrade to a more expensive hosting plan. This isn’t exactly comforting for a customer who’s looking at their declining visitor numbers while simultaneously seeing their infrastructure costs rise (unless they understand that AI crawlers are part of the reason).
Several organizations have already documented higher infrastructure costs with aggressive AI crawling:
| Website/Organization | AI Crawler Impact | Financial Impact |
|---|---|---|
| Read the Docs | One crawler downloaded 73 TB in one month | $5,000+ in bandwidth charges |
| Planetary client | Heavy scraping and repeated bot requests, including Anthropic traffic | 2x cloud computing costs |
| TechPays | AI crawlers caused ~30-40 GB/month in outbound data | $90 in data overage charges |
If one AI operator is repeatedly crawling thousands of pages while generating almost no referral traffic, while another sends meaningful visitors with comparatively little crawling, those are obviously very different traffic patterns, and potentially two very different costs for the customer.
Instead of looking only at how many people visited a site, they can look at the new bigger picture and answer: Where did those visitors come from? What did they do once they arrived? Which AI operators are crawling the site? And how much is that activity actually costing them?
