ChatGPT traffic 'drop' may be a tracking gap, not lost visits
SEO & AI search
A fall in ChatGPT-attributed visits this month looks more like a recording problem than a real loss. Check your analytics and your crawler settings before writing off AI search.
What has happened
Many marketers have opened their reports this month to find that ChatGPT is sending them less traffic. Search Engine Journal reports that ChatGPT-attributed traffic fell sharply in several analytics datasets. Analysts link the fall to ChatGPT links that open without tracking tags.
This explanation comes from analysts, not from OpenAI. So the fairest summary is that the drop appears to be a tracking problem. It may not mean fewer people are visiting.
The reasoning is simple. Analytics tools work out where a visit came from using two main clues. One is the referrer, which the browser passes along. The other is tracking tags added to the link itself. If a link from ChatGPT arrives with neither, your analytics cannot tell it came from ChatGPT. That visit still happens. It just gets filed somewhere else, often under "direct" or "unassigned".
Why this matters for your reports
If you only look at the ChatGPT line in your reports, a tracking change can look exactly like a loss of interest. That could push a business into bad decisions. It might cut content that is working, or decide AI search "isn't worth it" based on a number that no longer measures what it used to.
The better question is whether the total visits to the pages AI tools tend to send people to have changed. If your ChatGPT line has dropped but direct visits to those same pages have gone up, the visitors may simply have moved categories.
The second check: can ChatGPT read your site?
A tracking gap is one possible cause. A blocked crawler is another, and it is worth ruling out at the same time.
OpenAI runs separate bots for separate jobs. GPTBot collects pages for training AI models. OAI-SearchBot indexes pages so they can appear in ChatGPT search results. ChatGPT-User fetches a page when a person asks ChatGPT about it. Your robots.txt file can allow or block each one on its own.
Many sites block training crawlers on purpose, and that is a fair choice. The risk comes when a broad rule, a plug-in setting or a firewall also shuts out the search and user-triggered bots. Then ChatGPT may struggle to read or link to your pages, whatever your analytics say.
What this means for UK businesses
- Don't judge AI search on one line of a report. Compare ChatGPT, direct and unassigned visits to the same landing pages over the same period.
- Set up a channel for AI referrals. In your analytics, group visits from AI assistant domains together so they are easier to spot when they do carry a referrer.
- Check your server logs. They record bot visits, such as OAI-SearchBot and ChatGPT-User, that page-based analytics often miss.
- Read your robots.txt with fresh eyes. Make sure you only block the bots you mean to block.
- Check your firewall and security tools too. Bot protection can turn away AI crawlers even when robots.txt allows them.
Analysis compiled by Sayu (sayu.co.uk)
