The absence of referrers breaks the web’s social contract: according to Stox, AIs like ChatGPT and Gemini demand everything and give back little or nothing
Welcome to the eighth installment of SEO Confidential, our in-depth column featuring the key figures of international SEO. At a time when artificial intelligence is redrawing the foundations of search, data transparency has become the decisive battleground.
To analyze this scenario, we interviewed one of the most authoritative and technically skilled voices in the industry: Patrick Stox, Product Advisor and Brand Ambassador at Ahrefs.
With a background as a developer and solid experience at giants like IBM, Stox combines rare technical depth with a strategic view of the market.
In our conversation we tackled, with no filters, the most debated topic of the moment: the difficulty of tracking traffic coming from AI and LLMs.
And yes, whether it’s AI Overviews or ChatGPT, the story doesn’t change: data transparency remains a mirage.
An opacity that, according to Stox, hides data that is “not good” and breaks the pact of trust with content creators.
Here’s what Patrick told us. Enjoy the read.
“Google is not transparent with data. I think they know the situation isn’t great”, Patrick Stox
Patrick, Google confirms that traffic from AI Overviews will not be separated in Search Console. By now it’s clear they don’t want us to measure it, don’t you think?
Tracking is mixed in with regular search data and won’t be separated in Google Search Console, so yes, they don’t want us to be able to measure it easily.
In practice, when a user clicks a link that leads to your site, whether it comes from the new AI-generated answer (like AI Overviews) or from a “classic” search result, Google Search Console will record it as a single type of click. There will be no way to tell the origin apart. By merging the data, Google effectively makes it impossible for us to specifically analyze how our content performs inside AI answers.
Our readers are business owners who invest to be found online. If clicks from AI mode can’t be tracked, how can they tell whether their investments in SEO and content are working? Aren’t we at risk of making the wrong decisions in the dark?
The data available to us is becoming messier and less clear. This means it’s getting harder for a business owner to understand precisely which strategies are working.
If you invest in creating a high-quality piece of content, you won’t be able to know for sure whether the clicks it receives come from its strong traditional rankings or from the fact that the AI used it as a source.
This way we lose the ability to measure the return on investment (ROI) of SEO and content marketing activities in a granular way…
Exactly. This lack of transparency is not good news for anyone who owns a website and invests in their own visibility.
The question that naturally arises is precisely this: why did Google choose not to provide the data separately?
They’re being “shady” with the data. Precisely because they know how not-good it is.
Google says AI Overviews drive more traffic, but we can’t verify it. How do you do SEO like this, with the data kept hidden?
We simply have to keep doing what works: quality SEO. Even if we can’t measure AI Overviews results perfectly, this isn’t an entirely new problem. Many other marketing channels have some degree of “opacity” in their data, but that has never stopped us from using them successfully when the underlying strategy is solid.
Speaking of which, it seems clear to me that Google wants to keep users inside its ecosystem with AI answers. For those who don’t want to depend solely on Google Ads, what alternatives are there? Is there still room for effective strategies to generate traffic and sales outside these dynamics?
Right now, my strategy focuses on two fronts. On one side, I’m optimizing content to appear as much as possible in LLM answers and in those generated by Google’s AI (like AI Mode).
On the other, I’m working to diversify traffic sources so I don’t depend exclusively on traditional search. A concrete example is YouTube, which for many companies remains an exceptional channel for finding new customers.
Why, in 2025, are there still AI assistants that don’t pass referrers? Is it a technical oversight or a choice to hide the data? If Claude and Perplexity send them, why don’t ChatGPT and Gemini?
In my article I analyzed exactly this aspect. The main problem is that, in my view, few people actively check whether the referrer is being sent correctly. If nobody complains or measures the impact, for the developers of these systems it becomes a low-priority problem to fix.
This creates a paradoxical situation. On one hand, there are probably genuine bugs or rushed implementations causing the data loss. On the other, this lack of transparency breaks what I call the “social contract” of the web: we content creators allow bots to crawl our sites, and in exchange we expect to receive trackable traffic. If that doesn’t happen, the pact falls apart.
The even stranger thing is that it would be in their own interest to be tracked correctly. If an AI assistant shows up in marketing reports as a valuable traffic source, companies will start treating it as a strategic channel worth investing in.
The good news is that the situation is improving, perhaps precisely because people are starting to talk about it. ChatGPT, for example, recently fixed many of its attribution issues shortly after my article was published. I wonder if they read it…
More and more sites are blocking AI bots. After years of giving content away for free, are we now supposed to be happy that they use it to train models… without giving traffic back? Isn’t it time to demand fair value?
What you describe is not a future hypothesis, but a reality that is already taking shape. The most emblematic case is Reddit. Being an immense archive of human conversations, its content is invaluable for training artificial intelligence models.
Precisely for this reason, Reddit has already struck multi-million dollar deals to grant access to its data. So yes, the market is already moving toward content monetization, and the big publishers and major platforms are leading the way.
Are generic keywords now reserved for the big players? Does it still make sense to invest in them if users search directly for the brand?
Content quality is still the most powerful weapon. If you can create a resource that is genuinely superior to your competitors’, one that answers users’ questions excellently, you still have a chance to compete even on the most generic keywords.
That said, we can’t deny reality: the challenge is decidedly tougher for newer sites. Big brands start with a huge advantage in terms of authority and trust, factors that Google rewards. So yes, it’s possible, but it requires an exceptional investment in quality and more time to see results.
Does investing in advertising and backlinks really help you appear in AI Overviews? Or are we just chasing visibility in a system that rewards whoever is best known, not whoever is most competent?
Google still uses search signals for AI Overviews and AI Mode. Links, in particular, continue to be a key signal of trust and authority. Google uses these signals to understand which sources are the most competent and reliable. So yes, a good link profile absolutely helps.
If no one has an interest in creating original content anymore, where will AI get up-to-date information? And how will brands make themselves known?
I don’t think people will stop creating content, but perhaps less of it will be created. Now you have to decide whether there’s value and whether you want to distribute it publicly or privately.
I think companies will need to focus on multiple channels if they want to grow. YouTube, social media, etc.
In the era of AIOs and AI Mode, what happens to SEO tools? Do they still play a useful role in monitoring results, or are they in crisis?
I’m convinced the tools will eventually adapt. Meeting marketers’ needs is their reason for existing, so it’s only a matter of time before they offer solutions to measure this new scenario.
The new metric you can no longer ignore: getting cited by LLMs
If one single, brutal truth emerges from this conversation with Patrick Stox, it’s this: the “social contract” of the web, as he calls it, has been torn up. And not by accident.
His analysis leaves no room for doubt: when he calls Google’s attitude toward data “shady”, he’s not describing an accident, but a precise strategy. And as you’ve seen, with ChatGPT the situation is no different…
Your site is no longer a partner, but a data quarry for an AI that doesn’t return the favor. The era of trackable traffic, I’m sorry to say, feels like a distant memory.
Today, if you want to stay visible, your site has to speak a new language: the language of LLMs.
But that’s not all: you also need to diversify across channels like YouTube, for example, and keep investing in links.
Doing SEO today means building an ark while everyone else waits for the rain to stop. The dividing line, today, is this:
on one side, those who will keep optimizing for a black box, hoping for a report from Google that will never arrive.
On the other, those who build a brand so authoritative and multi-channel that it becomes the source the AI itself will be forced to cite.
Real SEO, today, is no longer about persuading Google’s algorithms. It’s about building a brand so authoritative that it is cited, mentioned and recognized as a primary source by artificial intelligences.
Well, we’ve reached the end of our article. All that’s left for me is to thank Patrick Stox for the pleasant chat, and you for your attention. See you next week for a new episode of SEO Confidential. I won’t spoil the guest, but get ready for big names!
#avantitutta
