We used to confirm; now we conform
In the 1940s, statistician Abraham Wald faced a problem that sounds strangely familiar today.
The military asked him to analyse bullet holes on planes returning from combat to decide where armour was needed most. The obvious answer seemed to be: reinforce where most planes were hit.
But Wald noticed the flaw — those bullet holes represented only the planes that made it back.
The data everyone relied on excluded the missing ones.
His insight: reinforce the areas without bullet holes — because that’s where the destroyed planes were hit.
That was an early lesson in confirmation bias and survivorship bias — drawing conclusions from visible data while ignoring what’s missing.
Fast-forward to today’s search ecosystem, and we’re doing the same thing.
We optimise and strategise based on the data we can see, without realising that what’s invisible might be far more important.
Only now, the issue has evolved.
Search isn’t just showing us partial data — it’s shaping what we’re allowed to see in the first place.
This is no longer just confirmation bias; it’s information bias.
Less visibility, less diversity
Recent changes have quietly deepened that problem.
We’ve lost the &num=100 parameter for seeing broader SERP data.
We get fewer query variations in Search Console exports.
And now, with Google’s new feature “Query Groups” in Search Console Insights, we’re being invited to see “similar queries grouped by AI intent.”
Sounds efficient, right?
It’s not — it’s reductive.
The messy, long-tail reality of search — where the most interesting user needs hide — is being replaced by smooth, aggregated intent clusters.
The problem? Those clusters reflect what Google thinks is similar, not what users actually express.
The new bias: data by design
When the data we use to make decisions is already summarised and filtered, our work becomes a mirror of that bias.
We start producing content for the “officially visible” intents — not the ones that fall through the cracks.
That’s how information bias is created:
- Google hides or groups away niche signals.
- SEOs adapt strategies to what’s exposed.
- The next generation of content reinforces those same patterns.
- Google validates the loop with more of the same data.
It’s a closed-loop bias:
we create based on what we’re shown, and what we’re shown is based on what we’ve created.
The illusion of more insight
The pitch for Query Groups is simple: less noise, clearer overview.
But what looks like clarity is often just loss of resolution.
If Google decides that “light guacamole recipe” and “easy guacamole dip” belong to the same group, the difference between health-conscious cooking and convenience cooking disappears.
And so does our chance to build meaningful, differentiated content for those audiences.
It’s not that the data is wrong — it’s that it’s been interpreted for us before we even touch it.
From search reflecting reality to defining it
In the early days, search engines tried to understand what existed on the web.
Now, they decide which version of the web we get to understand — and that has cultural consequences.
When visibility is algorithmically smoothed and human curiosity is mediated through AI summarisation, we end up optimising not for people’s needs, but for the system’s narrative.
The diversity of information — the raw, weird, non-standard queries — becomes collateral damage in the pursuit of cleaner metrics.
What SEOs can still do
We can’t stop the platform changes, but we can resist the intellectual laziness they encourage.
- Keep exporting and archiving raw GSC data before it’s aggregated.
- Cluster your own queries using NLP tools or regex patterns — your model, not Google’s.
- Use proprietary data — from CRM, analytics, or customer feedback — to uncover needs and language patterns that never appear in search data.
- Value originality. Search volume isn’t everything; the best-performing ideas often start with no visible demand at all.
- Value the outliers. The 0-click keywords, the odd phrasing, the things that “don’t fit.”
- Compare cross-source. Tools like Ahrefs, Brand Radar, or site-search logs reveal what Google no longer shows.
Real insight starts where Google’s grouping ends — and where our own curiosity begins.
Final thought
Every “simplification” hides a story.
We once struggled with confirmation bias — now we risk living inside information bias, where even our idea of reality comes pre-filtered.
If we let AI clustering decide what’s worth analysing,
we’ll keep seeing a cleaner picture —
but a narrower web.