Clicks No Longer Determine Your Rank. They Train the System That Decides Your Existence.
The Discomforting Paradox
For ages, the prevailing notion in SEO discussions was: “Clicks are unreliable. Google doesn’t place much emphasis on them.”
However, recent leaks, patents, and documents from antitrust trials have revealed a different narrative.
The 2024 Google API leak unveiled internal variable names that can’t be easily dismissed: goodClicks, badClicks, lastLongestClicks. The US v. Google antitrust trial made it clear – user behaviour plays a crucial role in refining rankings, and engagement is utilised to adjust results after they’ve initially been presented. Google’s internal terminology for its three ranking foundations? ABC: Anchors (links), Body (text), and Clicks.
So, it’s official: clicks do matter.
Yet, many stop their analysis here, missing out on a far deeper understanding.
Those signals don’t merely affect rankings; they also shape AI responses.
That’s the shift most founders and marketers haven’t fully grasped.
Key Takeaways
- The Discomforting Paradox
- What Cyrus Shepard Got Right (and Where More Nuance Is Needed)
- What Is Essentially True
- Where the Interpretation Falls Short
What Cyrus Shepard Got Right (and Where More Nuance Is Needed)
SEO expert Cyrus Shepard shared a thread claiming that click signals influence AI systems such as RankEmbedBERT, which in turn affects platforms like ChatGPT. While the general idea is on point, the mechanism is far more intricate than just “clicks equate to AI citations.”
What Is Essentially True
- Google employs implicit user feedback loops – including clicks, dwell time, and users bouncing back to the SERP
- Systems like RankEmbedBERT integrate search logs and human evaluations into their relevance models
- AI retrieval layers can build upon ranked outputs that were influenced by those behaviour signals
A more nuanced version of the assertion would be:
Clicks don’t merely rank content; they train the systems responsible for ranking content.
This is a vital distinction that alters how you should approach growth.
Where the Interpretation Falls Short
The assumption that follows – that clicks have a direct impact on ChatGPT – is somewhat misguided.
AI systems like ChatGPT:
- Do not rely on Google’s live rankings
- Do not access real-time clickstream data
- Do not optimise in real-time based on user clicks
What actually occurs:
They inherit the influence of click-optimised systems through training data and retrieval frameworks.
Subtle, yet significant. The influence is upstream and accumulative, not direct and instantaneous.
The True Model: Clicks as Upstream Training Signals
Let’s break down how the entire stack connects.
Layer 1: Clicks Influence Google Rankings
A high click-through rate (CTR) combined with prolonged dwell time indicates reinforced relevance. Conversely, poor engagement leads to demotion. This is well documented in Google’s patents and has been confirmed in court testimonies.
Layer 2: Rankings Affect the Web’s Content Layer
The content that ranks receives greater visibility, accruing more backlinks, mentions, and inclusion in listicles and reviews. The pages that excel compound their advantages over time.
Clicks determine what’s visible. Visibility determines what’s memorable.
Layer 3: The Web Becomes Training and Retrieval Input
AI systems learn from authoritative pages, frequently cited sources, and structured content that garners widespread reference. These are typically the same pages that triumphed in the click arena upstream.
Layer 4: AI Decides What Gets Cited
AI pulls from trusted, validated sources. It favours content already recognised by the broader ecosystem, amplifying what was selected upstream.
By the time AI cites your work, the click contest has already been settled.
Why This Matters More Than It Seems
This isn’t merely about CTR tricks. It’s about feedback loops shaping what information the public actually encounters.
The compounding cycle:
- Clicks bolster rankings
- Rankings elevate visibility
- Visibility enhances training data
- Training data fortifies AI responses
User behaviour doesn’t just shape results; it programmes the future of discovery.
Here are the implications that many haven’t fully considered:
- Early winners acquire disproportionate AI visibility
- Engagement becomes a data barrier that grows over time
- Mediocre content that attracts clicks can outlast genuinely superior content
AI doesn’t showcase the best content; it highlights the most reinforced content.
This represents a vastly different game from what most assume they’re playing.
My Perspective: Growth Involves Controlling Feedback Loops
Having developed growth systems across SaaS, Web3, and AI products, I recognise this pattern; the systems that succeed are rarely those with the best content at launch. They are the ones that manage the feedback loops long enough to allow compounding to take effect.
Clicks are no longer just performance metrics; they’re:
- Training signals – shaping what AI perceives as authoritative
- Reinforcement data – informing ranking systems about what truly satisfies users
- Selection bias at scale – determining which content receives amplification across the entire ecosystem
Growth is no longer about traffic; it’s about influencing the data layer from which AI learns.
This is where SEO, product, and distribution converge into a singular discipline.
What You Should Actually Do
It’s not about “optimising CTR.” That’s superficial. Here’s a more profound approach.
1. Design for Satisfaction Signals
Ensure fast load speeds. Offer content that addresses the query in the first paragraph. Maintain a clear structure and provide immediate value. Dwell time serves as a proxy for usefulness – and usefulness is what gets reinforced upstream.
2. Win the First Click Authentically
Create titles that precisely match intent. Avoid bait-and-switch tactics. Minimise pogo-sticking by delivering exactly what the headline promised. Winning a click but losing the user in ten seconds is far worse than not winning the click at all.
3. Engage in High-Engagement Ecosystems
Listicles, review sites, comparison pages – these are environments rich in behaviour where your content earns clicks in contexts that signal relevance to the algorithm. Being featured here feeds the loop in a way that individual blog posts simply can’t.
4. Eliminate Friction Relentlessly
Paywalls upon arrival, full-screen pop-ups, interstitials – all these factors increase bounce rates and create negative click signals. If users can’t easily access your content, they can’t engage with it, and you lose the vital signal you need.
5. Audit Your AI Visibility Signals
Beyond clicks, AI systems seek structured data, E-E-A-T signals, llms.txt, and crawler access. I developed SearchScore specifically to evaluate these signals – it runs a free audit in about 30 seconds, highlighting exactly where your gaps lie.
The Bottom Line
Cyrus Shepard is identifying something substantial. Yet the deeper truth extends beyond any singular thread.
Clicks don’t merely influence rankings; they mould the datasets that determine what AI accepts as truth.
And once something is deemed “true” within AI systems, it compounds. Those who recognised this early – who built for satisfaction instead of just traffic – are quietly constructing a data moat that will be incredibly challenging to breach.
The age of “get clicks” is behind us. We’ve entered the age of “earn the click.”
Let’s Keep in Touch
If this has sparked any thoughts – a new insight, a pressing question, a clearer signal – I’d love to continue the dialogue. I write about AI-augmented growth, systems thinking, and what’s genuinely effective at the cutting edge.
- Blog: ronniehuss.co.uk
- Medium: medium.com/@ronnie_huss
- LinkedIn: linkedin.com/in/ronniehuss
- X/Twitter: twitter.com/ronniehuss
- HackerNoon: hackernoon.com/@ronnie_huss
- Vocal: vocal.media/authors/ronnie-huss
- Hashnode: hashnode.com/@ronniehuss
