Key takeaway: ChatGPT sent far less traffic than Google in one study, while refreshed indexed pages corrected a Feature Upvote AI Overview.
Key Takeaway
Two reports published on October 21 describe separate but connected AI search and SEO issues. Marie Haynes examined 12 months of Google Analytics 4 data from 100 websites with more than 145 million sessions. ChatGPT referrals represented 0.62% of Google Organic Search traffic, while referrals from all included large language model tools reached 0.73%. Among the 55 sites tracking conversions, ChatGPT visitors converted at 13.8%, compared with 3.02% for Google Organic Search visitors. Google nevertheless produced 96.5% of the total conversions across the channels studied.
A separate Semrush report documented how Feature Upvote corrected a false Google AI Overview answer about its public REST API. Google had relied on outdated information indexed on the company’s website. After Feature Upvote updated the relevant pages and submitted them for indexing, the AI Overview returned a correct answer for some queries within about an hour. The result later varied by query wording before becoming more consistent.
What the ChatGPT and Google traffic data shows
The traffic study does not show that ChatGPT and Google are equivalent acquisition channels. ChatGPT visitors had a higher conversion rate in the measured sample, but Google delivered far more visits and therefore generated most of the conversions. The report said Google Organic Search produced 27 times more conversions than all LLM platforms combined.
For SEO and AI marketing teams, the distinction matters. Referral volume, conversion rate and total conversions answer different questions. A channel can send a smaller group of visitors who convert more often while still contributing less overall business value. The study also found that 45 of the 100 websites were not tracking any key events, limiting what rankings, mentions or citations alone can show about commercial performance. The report said ChatGPT referral growth in its sample slowed around August 2025, but that observation should not be generalized to every industry.
How Feature Upvote corrected a Google AI Overview fact
Feature Upvote’s case involved a specific brand fact: whether the product had a public REST API. Google’s AI Overview cited older blog posts saying it did not, instead of using newer help documentation. The company published and promoted an API announcement, corrected the outdated pages, and requested indexing for the relevant material.
The change was not uniform across every query. The report said an answer became correct about an hour after the indexing request for some searches, while another wording still produced the wrong response. It took several more days for the answer to become consistent. A later check on September 10 still returned “Yes,” although Google cited different Feature Upvote pages. Businesses monitoring AI search should therefore record exact prompts, answers and cited URLs rather than treating one successful check as a permanent fix.
Method and Limitations
Teams can use Google Search Console’s URL Inspection tool to check whether pages containing important brand facts are indexed. They can also keep a fixed set of prompts, record the wording and cited pages, and repeat the checks after updating content. In GA4, source, sessions, conversion rate and key events should be reviewed separately so that a high rate is not mistaken for a high total contribution.
The traffic findings are based on an observational sample of 100 websites, with conversion tracking available for 55, so they do not represent every business or industry. The Feature Upvote example is a single case. Updating pages and requesting indexing does not guarantee that an AI Overview will remain correct, and the reports do not establish that the same timing or outcome will apply to all brands, queries or search platforms.