Use Data Or Be Used By Data!

The October 12 issue of Seotistics is here for you!

Measurement did change in the last year and I am here to remind you some basic facts.

AI is for sure one of the main culprits but not the only one.

I am also writing a huge article on LLM measurement, stay tuned 👀

Hectic week, too much to do, including course updates.

Please move this email to your Primary inbox or reply to it. This is to prevent Seotistics goes into spam by accident. Gmail users can read this tutorial to do it.

Read this in your browser

📣📣 Important Announcements 📣📣

As for conferences/events, the next ones are:

SerpConf Rome - 16 October 2026 (this Friday!)
SEO On Stage (Athens) - 24 October 2026 (SEO-On-Stage-Marco70 for 70% off)
MeasureCamp Zurich - 21 November 2026

I've also decided to add a weekly news recap in every issue.

The first few times it will be simplistic but I am curious to experiment more here.

There will be new cool changes to this newsletter and plenty of content updates 👀

Become a Top Analyst Now!

News

From October 5 to October 12:

Experimental section, will develop it once I have more time 💀

If you ask me, the GA4 User Groups news is huge for programmatic work.

Anthropic also released some interesting stuff!

Bot Traffic & Issues

One of the main reasons why Web data is unreliable is often bot traffic.

GA4 requires quite some filtering work to be considered reliable, and what I usually do.

Instead of filtering from inside GA4, I recommend you actually do it from your data warehouse (or prevent bot traffic with a CDN).

I have a list of patterns I usually look for:

engaged_sessions at 0 across thousands of sessions, with engagement time under a few seconds
traffic landing on / only, with a single event per session
geographic clustering that makes no sense for the business (mine tends to arrive from China and Singapore)
device and browser fields that come back empty or absurd, including the odd 0x0 screen
one session per user and then nothing because bots don't come back
user agents that look like headless Chrome, python-requests or Go-http-client

GSC as a data source is ALSO affected by bot traffic, as we noticed from Google fixing and reducing impressions over time.

This can't be fixed unfortunately... and why I tell you to consider web data as directional.

Direct Traffic Nightmare

This is a big classic and something I also have to fix on my website 💀

Direct traffic is mistakenly defined as that bucket where all the people who visit directly land.

Imagine bookmarks or people reaching you via the search bar.

An example of why you need to care of your data quality, in the end linkedin could be merged and trust me, that direct can drop a LOT.

In practice, it's often related to technical issues you could prevent, like:

Links in email clients (especially desktop apps like Outlook), PDFs, and Word or Office documents
Messaging apps such as WhatsApp, Slack, Telegram, Signal, and SMS, which often strip referrers
Mobile apps that open links in in-app browsers without passing a referrer
QR codes
Social platforms that suppress or mask referrers in some contexts
HTTPS to HTTP transitions, which drop the referrer (and redirects)

The mindset shift is to stop treating Direct as one channel.

You can split it by landing page, device, new vs returning and consent status.

It can't by definition reach 0 but... you can at least control it.

Inside it, you may also have AI traffic!

OK.... But AI?

LLMs have been quite tough to classify because we relied mostly on:

page referrers
sources

The issue being in most cases that referrers are stripped (as just said) and we don't have much data from other tools.

AI traffic for now should be a simple indication/sample to evaluate potential conversions.

An example with ChatGPT, as you see anything goes.

Don't change your entire strategy for it, monitor what happens and make sure to report on that.

This is mostly for stakeholders. Marketing-wise, I don't even see the potential of this fixation.

Attributing Results to AI

You may have heard (or not) about incrementality as a way to measure the real impact of a given channel/source on revenue.

This is NOT easy at all because it requires companies to have a proper setup and system to track it all.

In the proper conditions, you could:

apply very specific changes to a page and measure what happens compared to a page without them
compare between specific regions/areas
compare with user groups

This is close to impossible in reality and most tests we read are Sci-Fi.

The big limit of Organic is that you can't switch it off (unlike Paid).

Well, technically you (partially) could by noindexing your website 💀