The Repetition Problem

What "formulaic" means in this research

A formulaic mention is a repeated wording pattern identified under the current repetition method.

The research first isolates exact or normalized duplicate structures.

Those repeated records are grouped into clusters so that several occurrences of the same formulation are not mistaken for several completely different pieces of language.

This matters because a campaign, publishing pattern or recurring recommendation can generate volume without generating the same amount of new information.

A repeated sentence still counts as a Brand Mention.

But it should not be interpreted as though it were a completely new formulation every time it appears.

What "unique-like" does not mean

The remaining 681 mentions are described as unique-like.

That phrase is intentionally cautious.

It does not mean organic.

It does not prove that the wording was independently written.

It does not prove that the mention was spontaneous, unpaid or created by a unique author.

It means only that the current repetition method did not classify the wording as an exact or near-text duplicate after the duplicate-detection process was applied.

That boundary is important because repetition analysis measures similarity.

It does not verify origin or intent.

The same mention count can describe two very different information environments

Imagine two brands with 100 mentions each.

For Brand A, 70 mentions repeat a small number of phrases.

For Brand B, the 100 mentions appear across many different questions, products, situations and experiences.

The headline number is the same:

100 mentions.

But the language environment is very different.

Brand A has high frequency and low diversity.

Brand B may have the same frequency but broader contextual diversity.

A count alone cannot distinguish those situations.

That is why repetition has to be measured separately.

Why repeated language matters for interpretation

Repeated wording can change what a researcher believes the dataset represents.

If the same formulation appears again and again, the dataset may contain fewer distinct ideas than the raw volume suggests.

That matters for several reasons.

First, repeated language can make visibility feel broader than it is.

Second, repeated formulations can dominate a dataset and make other contexts harder to see.

Third, semantic analysis can be distorted if the same relationship is reproduced many times through the same template.

The solution is not to delete repeated mentions.

The solution is to label them.

Frequency remains useful.

Repetition simply adds the missing context.

The Gold Dataset result

Within 744 explicit Brand Mentions:

63 exact formulaic mentions represent 8.5% of the explicit Brand Mention layer.

681 unique-like mentions represent 91.5%.

The percentages describe Gold Dataset v1 only.

They are not presented as the repetition rate of all 1,779,805 discussion URLs.

They also do not prove that 91.5% of the mentions are organic.

The result demonstrates something narrower:

a measurable portion of Brand Mention volume can consist of repeated wording, and that portion can be separated from the rest.

Why 21 clusters matter more than one repetition percentage

The 21 exact formulaic clusters show that repetition is not one single pattern.

Different wording formulas can repeat independently.

One cluster may represent a recurring recommendation phrase.

Another may repeat a product description.

Another may reproduce a similar company reference.

Looking only at the overall repetition percentage would hide those separate structures.

Clusters make it possible to study which formulations recur and how much of the volume each one contributes.

That becomes especially useful when comparing brands, campaigns or publishing periods.

Repetition also appears at the macro level, but it is a different measurement

The wider macro subset contains 1,493,085 offline threads used for structural title and category analysis.

Within that layer, the research identified:

• 39,203 exact duplicate-title groups
• 39,866 normalized repeated-title groups
• 156,339 threads in exact duplicate-title groups
• 168,663 threads — 11.30% — in normalized repeated-title groups

These figures are not Brand Mention repetition counts.

They describe repeated discussion-title structures.

The two layers should not be merged.

Gold Dataset repetition measures repeated wording inside explicit Brand Mentions.

Macro repetition measures repeated title framing across a much larger thread set.

Both show that repetition exists.

They measure different things.

Why this matters for Brand Mentions

A Brand Mention can create value even when similar wording already exists elsewhere.

The point of repetition analysis is not to declare repeated mentions worthless.

It is to understand what kind of volume is being produced.

A campaign with 1,000 repeated formulations and a campaign with 1,000 contextually diverse mentions are not the same information environment.

If both are reported only as "1,000 Brand Mentions," that difference disappears.

Repetition analysis puts it back.

Why this matters for AI Search research

AI systems can encounter repeated public information across many pages.

The current study does not claim that repeated Brand Mentions are rewarded, ignored or penalized by AI systems.

That question requires a separate experiment.

What the current research establishes is the measurement layer needed before such an experiment can be meaningful.

If repeated and non-repeated mentions are mixed together, it becomes difficult to test whether diversity, persistence or frequency is associated with different AI outcomes.

Repetition therefore matters first as a data-quality question.

Only later can it become an AI Search question.

What this study does not claim

The study does not claim that formulaic language is spam.

It does not claim that repeated mentions are artificial.

It does not claim that unique-like mentions are organic.

It does not claim that more diverse language automatically creates more value.

It does not claim that repetition causes or prevents AI recommendations.

The finding is narrower:

Brand Mention volume and Brand Mention diversity are not the same metric.

That distinction is the result.

The practical takeaway

When a report says:

"This brand received 5,000 mentions,"

the next useful question is:

How much of that volume repeats the same language?

Then:

How many distinct formulations appear?

How many repetition clusters exist?

Do repeated mentions dominate the total?

Does the same semantic context repeat, or do the mentions introduce new products, problems and use cases?

Without those questions, volume can be mistaken for diversity.

Related research

Not All Brand Mentions Are Equal
Why mention type also needs to be separated before volume is interpreted.

A Link Is Not Always a Brand Mention
Why links and explicit naming are different signals.

Brand Mentions Before the Decision
Why conversational role matters alongside volume.

Formulaic Language v1
The derived dataset behind the repetition findings.

Research Limitations
The boundaries that apply to the repetition analysis.