If dozens of credible webpages say roughly the same thing and one page says something different, which information is an AI search engine more likely to use?
At first glance, the answer seems obvious.
The majority view has weight. There are more sources supporting it, more repetition around it and more evidence that the conclusion is established.
But there is another force working in the opposite direction.
If those sources all repeat substantially the same information, another version of the consensus may add very little.
A page that contributes a useful distinction, original evidence or specialist knowledge may provide something the existing sources do not.
That creates an interesting tension:
Consensus has strength through repetition. Distinctive information has value through uniqueness.
Our recent Google AI Mode experiments gave us a practical example.
A Strong Existing Consensus
We asked Google AI Mode how many times someone should repeat the same query when testing whether their business was consistently recommended.
Before publishing anything new, we ran the identical query five times.
The result was:
30 / 30 / 30 / 30 / 30
AI Mode repeatedly treated 30 as something close to a statistically reliable minimum.
We then published a resource explaining why there is no universal number that automatically makes every AI visibility test statistically reliable.
Our argument was not:
Everyone says 30, but we say five.
It was:
A quick repeatability check and a formal statistical estimation exercise answer different questions, so they should not automatically use the same methodology.
Google indexed the page quickly.
We then repeated the broad query another ten times.
The new resource was cited:
0 times out of 10.
Thirty remained the dominant answer, although the recommendations occasionally dropped to 10 or 15.
When we later made the user’s intent more specific and explicitly asked for a quick non-statistical consistency check, the answer changed sharply towards five runs.
Eventually, when we asked AI Mode to find a resource explaining that exact distinction, it retrieved and cited our page.
The full experiment is covered separately, but the useful question here is:
What happens when your distinctive information conflicts with an established consensus?
Why Consensus Is Powerful — and Why It Can Still Create Opportunity
Consensus deserves weight when good sources independently support the same conclusion.
A page does not become valuable simply because it disagrees.
Suppose many reputable accounting sources say cloud accounting is suitable for most small businesses.
One article recommending desktop software does not suddenly become more useful merely because it is different.
But a specialist page explaining that certain businesses with legacy integrations, unreliable connectivity or unusual operational requirements may need a different setup could add something important.
The distinction is not:
majority versus minority.
It is:
Does the minority view contribute information the majority answer does not adequately cover?
That is where distinctiveness becomes useful.
Commodity Consensus Creates Substitutes
There is another problem with very well-established information.
Suppose 100 pages all explain the same concept in almost identical terms.
The consensus may be strong.
But from a source-selection perspective, the pages may also be highly interchangeable.
What does page 101 add?
Potentially very little.
That creates a practical problem for websites trying to become visible in AI-generated answers.
If your page gives the AI essentially the same information available from dozens of other sources, the system has many substitutes.
That leads to a principle we have been developing through our testing:
If your information is interchangeable, your webpage may be interchangeable as a source.
The opportunity is therefore not necessarily to contradict consensus.
It is to contribute something the existing answer genuinely needs.
Minority View or Minority Circumstance?
This distinction may be even more useful.
Sometimes you are not really saying:
The majority is wrong.
You are saying:
The majority answer is right for most people, but this customer’s circumstances place them in an exception.
That is a very different kind of content opportunity.
An idea can simultaneously be:
- a minority view across the broad topic;
- the most appropriate answer to a narrow situation.
Suppose approach A is suitable for 95% of customers.
Approach B only makes sense for 5%.
For a broad query, A should probably dominate.
But when the user describes the circumstances defining that 5%, B may become the best-fit answer.
That is not contrarianism.
It is specificity.
Query Specificity Can Change Which View Becomes Useful
We saw this clearly in our own AI visibility testing.
The broad question:
How many times should I repeat an AI visibility query?
strongly gravitated towards 30.
The narrower question:
I am not trying to estimate a statistically significant probability. I just want a quick manual check to see whether the same Google AI Mode query gives consistent recommendations.
produced:
5 / 5 / 5 / 5 / 3–5
The topic had barely changed.
The user’s purpose had.
That changed which information was most useful.
The broad query invited general statistical and monitoring advice.
The narrow query removed the need for formal statistical estimation and made practical repeatability methodology much more relevant.
That suggests a useful principle:
A minority idea does not necessarily need to overthrow the broad consensus. It may only need to become the best answer when the user’s circumstances become specific enough.
Distinctive Does Not Mean Contrarian
There is an important difference between these two approaches.
Contrarian content
Everyone says X. I say Y.
That may be unusual, but there is no reason to trust it.
Evidence-backed distinctiveness
Most sources say X, but that conclusion combines two different situations. Here is the distinction, here is the evidence supporting it, and here is when the alternative matters.
That is a genuine contribution.
Our 30-search example fits the second category.
We were not arguing that 30 searches are inherently wrong.
We were arguing that:
A quick repeatability check and a formal statistical study answer different questions.
Five runs were simply a practical starting point for the first purpose.
The useful information was the distinction.
The Further You Move From Consensus, the Stronger the Evidence Should Be
This may be the most important practical rule in the whole discussion:
The further you move from consensus, the stronger the evidential burden becomes.
If you are making an unsurprising claim that is widely established, the supporting burden may be relatively light.
If you are challenging a familiar rule, you need to show why.
Useful evidence might include:
- original experiments;
- first-party data;
- specialist experience;
- transparent methodology;
- recognised external research;
- calculations;
- documented exceptions;
- real case studies.
Being different gets attention.
Evidence gives the difference weight.
First-Party Evidence Creates a Different Kind of Distinctiveness
Some of the strongest distinctive information does not involve disagreement at all.
It may simply involve facts only your organisation can provide.
A dental practice can explain exactly how it treats nervous patients.
An accountancy firm can explain exactly how it works with manufacturers.
A hotel can explain exactly how it accommodates a specific accessibility requirement.
A software company can explain exactly how a particular integration behaves.
That information is distinctive because it is genuinely first-party.
Compare:
We work with manufacturers.
with:
We provide monthly management accounts for manufacturing businesses using Xero, including reporting around stock, margins, payroll and cash flow, on a fixed monthly fee.
The second gives a customer much more evidence about whether the firm fits their situation.
It also gives an AI system more concrete information to work with.
Another business cannot simply manufacture those facts about your service.
That makes the information less substitutable.
Sometimes the Consensus Is Not Wrong — Just Too Simple
There is another content opportunity worth separating from straightforward disagreement.
Sometimes the facts exist.
The problem is the interpretation being placed on them.
Our 30-search experiment is a good example.
The underlying ideas around sample size, repeatability and AI variability were not imaginary.
The questionable step was compressing them into the broad rule:
You need 30 searches for an AI visibility test to be statistically reliable.
The facts existed.
The interpretation was too simple.
We think this may represent a separate opportunity we call an Interpretation Gap:
the information exists, but the prevailing conclusion drawn from it overlooks an important distinction.
That deserves a post of its own.
For now, the important point is that useful content does not always require finding a new fact.
Sometimes the opportunity is explaining existing information more accurately.
Before Challenging the Consensus, Ask Three Questions
You do not need to actively search for things to disagree with.
That would encourage poor content.
Instead, ask:
1. Is the consensus actually wrong, or merely incomplete?
Sometimes the standard answer is perfectly adequate.
Sometimes it simply fails to cover an important circumstance.
2. What specific condition changes the answer?
What makes your exception relevant?
That might be:
- industry;
- business size;
- location;
- budget;
- previous experience;
- urgency;
- technical constraints;
- software already in use;
- unusual customer requirements.
If you cannot define the circumstance, the disagreement may be too vague to be useful.
3. What evidence supports the alternative?
Can you demonstrate why the standard answer stops being sufficient?
If not, being different adds very little.
Do Not Assume Joining Information Together Makes Your Page Essential
There is another complication.
AI systems can synthesise information from several sources.
We saw this during the same experiment.
When we narrowed the query to a quick non-statistical repeatability check, AI Mode produced methodology very similar to our new page without citing us.
It had enough information elsewhere to construct the answer.
That means even a well-organised resource may remain substitutable if AI can easily reconstruct its individual claims from other sources.
The stronger opportunity may therefore be to combine good synthesis with something harder to replace:
- your own evidence;
- your own testing;
- your own data;
- your own first-party knowledge;
- a genuinely useful distinction.
The question is not simply:
Is my page complete?
It is:
What does my page contribute that would be difficult to reproduce without it?
A Practical Question for Businesses
When reviewing content, ask:
If AI already has 50 sources saying this, why would it need mine?
There may be a good answer.
Perhaps your page contains original research.
Perhaps it explains a specialist circumstance.
Perhaps it documents how your own service actually works.
Perhaps it provides better evidence.
Perhaps it identifies an exception other pages overlook.
Perhaps it explains the customer’s problem more completely.
But if the answer is:
Mine says essentially the same thing in slightly different words.
then the resource may be highly interchangeable.
That does not make it useless to human readers.
It simply gives an AI system relatively little reason to select that particular source.
What Our Experiment Does — and Does Not — Show
We should not turn these observations into claims about Google’s internal ranking systems.
We have not proved that Google counts consensus sources.
We have not proved that distinctive information receives special weighting.
And we have not proved that publishing a minority view improves AI visibility.
What we directly observed was more limited:
- a broad query initially produced a perfect five-run consensus around 30 searches;
- our new dissenting resource did not appear in ten subsequent broad-query tests;
- narrowing the user’s intent changed the dominant answer dramatically;
- increasingly specific queries produced reasoning similar to our page without citing it;
- when we explicitly asked AI Mode to find a resource addressing our distinctive distinction, our page was finally retrieved.
Those observations are consistent with the idea that the usefulness of distinctive information depends heavily on the particular information need expressed by the query.
That is the hypothesis worth exploring further.
The Bigger Principle
Consensus is valuable when many strong sources independently support the same conclusion.
Distinctiveness is valuable when it adds something those sources do not.
The opportunity is therefore not to disagree for attention.
It is to identify where the standard answer stops being sufficient, explain the circumstance that changes the answer and support that distinction with evidence.
Do not disagree merely to be different. Do not agree merely to blend in.
Understand what is already known.
Find what is genuinely missing.
Support what you add.
Because in an information environment where AI can already access dozens of versions of the standard answer, the most useful contribution may be the one that gives the answer something it did not have before.