Spike in Direct Traffic vs SEO

Ooo this is an interesting one!

Are your Direct (iOS) visits landing on the same types of pages as your SEO traffic on other devices? If yes, that’s usually a bit of a giveaway that the traffic is still coming from search, it’s just no longer being attributed as such.

The way Adobe works, if the referrer isn’t there, it can’t classify the visit as Natural Search, so it falls into Direct. Unfortunately, if iOS is removing the referrer before the data reaches Adobe, then Adobe simply doesn’t have the signal it needs, so there’s no straightforward way to reclassify that traffic without implementing some potentially damaging rules.

If you want to sanity check what’s happening, it’s worth following the journey yourself and inspecting the referrer. For example:

• Open the dev console and check document.referrer when you land on the site
• Or use the Network tab to inspect the request headers and see if a referrer is being passed

If you’re coming from a Google result on iOS and the referrer is already missing or empty at that point, then Adobe never receives that signal, which explains why it ends up in Direct. It’s a simple test, but a really useful way to confirm whether the referrer is being dropped before it even reaches your analytics.

The other thing that stands out to me is that it’s only happening on one project. If this were purely an iOS change, you’d expect to see it across other markets/clients/projects too. That usually means iOS is part of the story, but something in this specific setup is amplifying it. A few things I’d look at:

• Marketing Channel processing rules
  If the rules are stricter here, anything with a missing or partial referrer will fall straight into Direct, whereas other setups might still recover it as Natural Search
• Redirects or cross-domain behaviour
  Additional hops (HTTP to HTTPS, tracking domains, JS redirects) can cause referrer loss, and iOS tends to be more sensitive to that
• Referrer policy settings
  A stricter policy could mean the referrer is being dropped more aggressively on this site
• Consent / CMP timing
  If Adobe fires after consent and there’s any delay, the referrer might be lost before it’s captured

So it’s less likely that iOS is behaving differently just for this market, and more likely that this implementation is less resilient to the referrer loss that iOS introduces. Is there anything standing out that’s different? 

Fully agree with what ​@LizzieSc wrote!
 

Safari has its advanced fingerprinting protection which protects the users from various invasive fingerprinting techniques, one of which is removing referrer information. But this does not explain the drop only on Spanish pages.

Are you certain that your Spanish campaign does not get redirected and hence may lose the referrer information e.g. through a URL shortener?

But then again, this would also affect Android traffic.

Would you mind sharing the URL where you see the behavior in a private message for further investigation?

That is very interesting…

As ​@LizzieSc said, I think the best way to see what is happening is to test it yourself. You may or may not be aware, but proxy testing (a common method of testing mobile apps before AEP Assurance / Griffon) can also be used to test your web data on your mobile devices.

This way, you can use real iOS and Android devices, perform a search, and click through to your website… then you can check the referrer that is passed. Since this is Natural Search, there won’t be any campaign codes, so your channels will rely solely on the referrer.

If the new iOS is dropping the referrer, then we are all in trouble… but if it’s not, you will need to check your Paid/Natural Search rules, and your Marketing Channel rules to make sure that nothing was missed there.

Good luck, and let us know how you make out.

Thanks to everyone who replied :)

Talking to our HQ about this case, and I have some updates.

1. Apparently, this is happening in other countries, not only in Europe/EU (which I assume could be the case), but also in the UK, Australia, Finland, and Germany confirmed their Adobe is reporting a spike in Direct and a drop in SEO, also segmenting for iOS 26 + versions users.
2. On the other hand, this is only happening in my project. HQ has another project within the same niche, and they’re not experiencing results like this.
   1. We considered the % of iPhone users on both, and they are kinda balanced - my project has around 60% iPhone users, while the other project has 48%.
3. I’ve checked other channels, and this is only happening with Direct and Natural Search. On the other hand, Paid Search shows growth and a trendline change around the same time (january), but there was also a budget increase for PS campaigns, so this explains it.
4. We checked Google Search Console, and it’s reporting the same drop.
5. I still can’t check because in my office I work with windows hahahah so as soon as I go home, I’ll try and check the document.referrer. on my apple products.

I also found this article explaining Apple’s constant updating Safari’s privacy protection and the new AFP (Advanced Fingerprinting Protection)

https://medium.com/billy-grace/safari-on-macos-ios-26-tracking-changes-whats-really-changing-31e2d26cb727

Anyways, any other updates, I’ll let you know.

Thank you so much!

If the Referrer is lost, the data in the Referring Domains channel will also decrease.

We had a similar issue with one of our B2C website, and the root was cloudflare bot detection screen stripping off referrer, technically ‘no referrer’ presence whereupon website’s landing page thus Adobe reported direct traffic is not unexpected. Suggested cloudflare managed page embed/dispatch the Adobe hit, so visit capture little advanced to ensure referrer comes into Adobe before landing page; compromise slightly BR getting affected, but BR metric to adjustable in Adobe luckily. Testing the journey by yourself may uncover gaps like this, which might be unique. Thanks

Hi ​@carolmadureira 
I’m just following up to check if any of the responses helped address your query. If so, please consider marking it as the best answer.