Google is not confirmed to be testing tracking parameters in AI Overview links. A parameter, ?st_source=ai_overview, was spotted on links in AI Overviews and AI Mode and reported by Search Engine Roundtable as a possible Google test, but the article was later updated: it turned out a Chrome browser extension was modifying the URLs. Google has said nothing about it in the source I read. Facts are as of 4 October 2026.
Key Takeaways
- Tom Pool shared screenshots on LinkedIn of links carrying
?st_source=ai_overview; Barry Schwartz could not replicate it across browsers. - Search Engine Roundtable's correction says a Chrome extension was adding the parameters, "presumably to sell the data on to brokers," so this was not a Google test.
- Google's documentation says AI feature traffic is reported in Search Console's Web search type, not separately.
- A clean URL parameter would be useful; do not build reports on one that does not exist.
- A privacy lesson: check browser extensions on the machines you use for testing.
What was reported, step by step
- The sighting. Per Search Engine Roundtable, Tom Pool posted screenshots showing a parameter,
?st_source=ai_overview, appended to links within AI Overviews and AI Mode. - The replication attempt. The article's author tried several browsers and could not reproduce it.
- The correction. An update states it was a Chrome browser extension modifying URLs to include the parameters, with the suggestion that the aim was to sell data to brokers. That is the publication's wording; I have no independent evidence of motive.
- Google's role. The article contains no Google statement.
The headline framing of a "Google test" is therefore wrong, and this post exists partly to say so clearly. Some third-party articles may still describe it as a Google experiment; check dates and corrections.
Why this still matters for attribution
AI Overview and AI Mode clicks are hard to separate from ordinary search clicks. Google's documentation on AI features states that traffic from AI features is counted in the Search Console Performance report under the Web search type, and it adds that clicks from results pages with AI Overviews are "higher quality" (Google's claim, not independently verified here). Nothing in the documentation I read describes a distinct URL parameter for AI links.
That leaves you three realistic measurement routes. They are all imperfect.
Route 1: Search Console, query and page patterns
You cannot filter to AI Overviews directly, but you can watch pages and query classes that tend to trigger them. My explainer on the Search Console generative AI report covers the generative AI report some sites can see and what it does not show.
Route 2: Referral sources for AI assistants
Chatbots that send visitors, such as ChatGPT or Perplexity, often appear as referrers in analytics. That is a different channel from Google's AI features, so keep it in a separate segment.
Route 3: Controlled prompt testing
Run a fixed query set weekly, record whether your page is cited, and compare with traffic. It is manual, but it is evidence you own. My checklist is in why a competitor ranks in AI answers.
What a real Google parameter would change
If Google did add an official parameter, analytics teams could build a GA4 segment from utm-style or custom query values. That would help in four ways:
- Separate AI-sourced clicks from classic organic.
- Compare conversion rates.
- Measure the effect of citation changes, such as the one I describe in the Gemini 3.8 Flash citation bug post.
- Make publisher discussions, such as the AI contribution pilot, more data-based.
For context on why this matters now: Google's Gemini 3.8 Flash bug showed links in AI Mode can vanish for a day, and the AI contribution pilot reporting shows publishers cannot audit what Google does with AI exposure. Better attribution data would help on both counts. Google's guidance also keeps pushing quality over mechanics: its helpful content guide and its generative AI content guidance are the documents to follow while measurement stays imperfect.
It would also carry trade-offs: more visible AI traffic share could encourage more scrutiny of click-through declines. None of this has happened. Google's own pages list no such parameter, so treat any report to the contrary as unverified.
A quick checklist if you see odd parameters in your logs
| Check | Why |
|---|---|
| Does it appear in your server logs or only in screenshots? | Extensions change what a browser shows or sends, not necessarily what Google serves |
| Does it appear across several clean browsers and devices? | A real test would |
| Is the parameter name tied to a third-party tool? | st_source is not documented by Google in anything I opened |
| Did a reputable source confirm with Google? | Not here |
| Is your own tag manager adding it? | Surprisingly common |
I would also run the test in a fresh browser profile with no extensions. That is the single most useful step.
Privacy and security angle
If a Chrome extension is rewriting links in search results and sending data elsewhere, that is a security issue for whoever installed it. For teams, I would:
- Audit extensions on work machines used for SERP testing.
- Remove anything you do not recognise or need.
- Keep a "testing" profile with no extensions for SEO and AEO checks.
Google's own advice on avoiding spam and manipulation is in its spam policies; extension-driven data selling is a separate matter from those policies, and I am not suggesting otherwise.
How I would set up AI traffic reporting today (as of 4 October 2026)
- Define three buckets: Google organic (including AI features, not separable), AI assistant referrals, and branded search.
- Add a monthly manual citation audit of 20-50 priority queries.
- Track brand search impressions in Search Console. A rise alongside stable clicks can hint at influence without a click.
- Use annotations for model and interface changes so you can explain dips.
For the interface changes this month in particular, see my post on Gemini 3.8 Flash and visibility.
Why rumours like this spread
Attribution for AI features is a genuine pain point, so a screenshot that appears to solve it travels fast. The pattern is familiar: one person sees something odd, a news site reports it with a question mark, social posts drop the question mark, and a correction arrives days later to a smaller audience. My rule is to wait for one of three things before changing a report: reproduction on a clean setup, an entry in Google's documentation, or a named Google representative. Here none arrived, and the correction pointed the other way.
The same discipline applies to other AI search claims you will see this month. Ask who measured, with what sample, on which date. If the answer is a single screenshot, label it an observation, as I did with the Gemini 3.8 Flash bug, where no citation-rate data was published at all.
What would change my mind
I would revise this if Google updated its documentation with an official parameter, if Search Console added an AI filter, or if Roundtable reported a Google confirmation. Watch Google's Search Central updates log, which on 1 October listed a revised generative AI content guidance page but no tracking parameter change.
FAQ
Is Google adding tracking parameters to AI Overview links?
There is no confirmation. A parameter, ?st_source=ai_overview, was reported, but Search Engine Roundtable later updated its article to say a Chrome extension was modifying the URLs.
Who spotted the parameter?
Tom Pool shared screenshots on LinkedIn, per Search Engine Roundtable.
Could Barry Schwartz reproduce it?
No. He tried multiple browsers without success, and the correction followed.
Where do AI Overview clicks show in Search Console?
Google's documentation says traffic from AI features is included in the Web search type in the Performance report. There is no separate AI Overviews filter in what I read.
How do I measure AI Overviews traffic then?
Use Search Console pages and queries likely to trigger AI features, track impressions and brand search, and run a repeatable manual citation audit. None of these is a perfect isolate.
Should I strip unknown parameters from my analytics?
Do not strip by habit. Investigate first. If a parameter comes from an extension, it will not appear for most users, so it will not skew your data much but may confuse test screenshots.
Is the extension a security risk?
Search Engine Roundtable's note suggests it may be collecting data to sell. If you use an unknown extension that rewrites links, remove it and review what it could access.
Will Google ever add an official parameter?
Unknown. There is no announcement in the documentation or news I checked.
Work with me
I'm Younus Fardeen, an organic growth, SEO and AEO marketer with 4+ years of marketing experience in edtech and startups, including Masai School's social growth (Instagram 26K to 117K, LinkedIn 50K to 200K). If you want AI traffic reporting that does not rely on rumours, see my work and use the contact form at younusfardeen.in.