Oops– a hypo**

I was giddy when popped the URL for Microsoft's new "How AI Search Is Changing the Way Conversions are Measured " into my Rich Snippets working doc. I was a bit annoyed when I had scrolled halfway through the article to get past the BIG IDEAS and BIG PROMISES to get to the heading I had been waiting for: "Does AI Search Traffic Convert?"

"The latest research from multiple independent sources points to the same trend: AI-search visits may be smaller in volume but are producing more qualified traffic and higher conversion rates than traditional organic search.

Amsive found that 56% of sites saw higher conversions from AI-driven sessions, with high-traffic sites converting at 7.05% compared to 5.81% for organic3. Similarweb reported a similar pattern, with AI referrals converting at 11.4% versus 5.3% for organic across global ecommerce4."

I click on the Amsive link and recognize the title "Does LLM Traffic Convert Better Than Organic? A New Data-Backed Study". It was the first piece I'd read by Wil Guevara and I loved everything about it. I took particular petty delight in the study's key finding:

"When looking at averages across all sites, organic traffic converted at 4.60% while LLM referrals converted at 4.87%. At first glance, this suggested a modest advantage for LLM traffic."

Well, goodness.

I'm sure Microsoft would never misrepresent data to give the appearance that LLM traffic converted at a higher rate than orangic. Citing a select flattering subset to represent "Does AI Search Traffic Convert?" without providing the important– and likely more applicable to the reader– subset is certain a picked choice. A cherry of convenience!

Someone should send over the full study so the blog can be corrected to show the full and robust data! I mean…. It's the only one linked in the body. I had to go to the foot notes to find SimilarWeb report.

Or rather– SimilarWeb Press Release. (PR professionals are renowned for their work in statisical analysis, afterall). The citation is actually a TL;DR presser but if you follow that link then signup to get a copy of the actual report from the landing page– after providing your job title from the drop down– you'll land on a 62 page mutlimedia report that you CAN'T SEARCH.

It's almost like they're trying to make it hard to verify the claim.

…

Someone should let Microsoft know so they can include more data points, update that link to the primary source, and not accidently create unfounded hype.