Practical AI Learning Series • Guide 13

The Template You Already Recognise

Why AI output looks the same every time, and what that sameness replaces

1. You Have Read This Before

Read the following paragraph:

AI-Generated Output — Example A “This is not a technology problem. It is a governance problem. Every major institution is built on assumptions that no longer hold. Every regulatory framework assumes human-speed decision cycles. Every hiring process filters for signals that AI has made meaningless. The question is not whether these systems will break. The question is whether we notice before the damage becomes structural.”

It sounds insightful. It sounds like someone who understands the territory. If you received it from an AI tool after asking for analysis of your organisation’s AI readiness, you might feel you had been given a thoughtful, provocative answer.

Now read this:

AI-Generated Output — Example B “This is not a marketing challenge. It is a positioning challenge. Every successful brand in the next five years will be built on principles that most agencies do not yet understand. Every campaign strategy assumes consumer attention patterns that have already shifted. Every measurement framework tracks metrics that no longer predict outcomes. The question is not whether your current approach will fail. The question is whether you adapt before your competitors do.”

Same structure. Same rhythm. Same rhetorical moves. Different topic. Both could have been generated by any major language model in response to almost any strategic question. Neither says anything specific enough to be wrong, which also means neither says anything specific enough to be useful.

Key Principle This is not a coincidence. It is a template. And understanding why it exists, what it replaces, and how to respond when you see it is one of the most immediately practical skills you can develop for working with AI.

2. The Template, Mapped

2.1 The Four Moves

The default AI rhetorical structure follows a consistent pattern. Not every AI output uses all four moves, but the pattern is recognisable enough that most experienced AI users will have encountered it dozens of times:

The Opening Negation

“X is not Y. It is Z.” This creates the sensation that a common misconception is being corrected. In genuine analysis, this positioning is earned through engagement with competing claims and evidence for why the conventional framing fails. In the template, the negation is asserted. The conventional view is implied but never examined. The reframing arrives without justification.

The Staccato Facts

A sequence of declarative statements, each given its own sentence for emphasis. These feel like evidence because they are formatted like evidence: separated, specific-sounding, building in sequence. But separation is not substantiation. A series of assertions arranged vertically is not a chain of reasoning. It is a list presented as though it were one.

The Analogy Pivot

A shift to a new metaphor or conceptual frame. This performs the structural role of analytical reframing: the moment in a genuine argument where the author shows the reader a new way to see the problem. In the template, the reframing is purely rhetorical. The new frame arrives with confidence rather than reasoning.

The Confident Close

Implications delivered with certainty. This performs the role of a reasoned conclusion, delivered with the authority of someone who has completed a rigorous argument. Except the rigorous argument never happened. The template reached its conclusion at the negation. Everything between was performance.

2.2 Why This Template Specifically

A reasonable question: why does AI default to this particular structure and not some other one?

The answer connects directly to how these models are trained. As Guide 02 (The Sycophancy Problem) explained, models learn what to produce by being trained on human preference ratings. Raters working quickly, evaluating outputs they often lack the domain expertise to verify, respond to structural markers of quality rather than to quality itself. This template is dense with those markers: it signals intellectual authority (the negation), evidential rigour (the staccato facts), analytical depth (the reframing), and confident expertise (the closing implications).

Important Distinction The template was not invented by AI. It was already the dominant register of LinkedIn thought-leadership, TED-style presentation, and management consulting output. Human content creators had already optimised it for engagement. RLHF training identified it as what humans rate highly, and the models industrialised it. They produce it not because it is good analysis but because it is what good analysis looks like to someone evaluating quickly.

3. What the Template Replaces

This is where the observation becomes practically important. Each element of the template performs the structural role of a genuine reasoning operation without actually executing it. Understanding the substitution is what lets you evaluate whether a given piece of AI output contains real analysis or just the shape of it.

Template Element	What It Performs	What Genuine Analysis Requires
Opening negation	Positioning against existing views	Citation, engagement with competing claims, explanation of why the conventional framing fails
Staccato facts	Evidential development	Sourced evidence, causal reasoning, acknowledgement of limitations
Analogy pivot	Analytical reframing	Identification of cases the prior frame cannot accommodate, reasoning about why the new frame succeeds
Confident close	Reasoned conclusion	Conclusion that follows from the preceding argument, with appropriate qualification

The framework I have been developing across the Mirror Effect essay series calls this proxy collapse: the condition in which a signal that once correlated with quality becomes dissociated from it. Production difficulty used to proxy for competence. Producing articulate, well-structured analysis was hard enough that doing it at all was evidence you could probably also think. That correlation broke when generation became cheap. The writing template is the same collapse at a finer resolution. The structural markers of good reasoning used to correlate with good reasoning because producing them required actually doing the reasoning. Now they can be generated without it.

Guide 03 (Verification vs. Generation) made the point that generating output and verifying output are different skills with different costs. The template is the most visible example of that gap in operation: the generation looks like verified reasoning, but no verification has occurred.

4. The Detection Exercise

Try It Now

Take a piece of AI-generated analysis you have received recently, something you asked for in the course of your work. Read it through and check for the following:

Opening negation or reframing. Does it begin by correcting a misconception? If so, does it say who holds the conventional view and where specifically it breaks down? Or does the new position simply arrive?
Declarative sequences. Are there passages where several assertions appear in sequence, each formatted for emphasis? Are they sourced? Could you verify them independently?
Analogy or metaphor shifts. Does the analysis pivot to a new conceptual frame? Does it explain why the previous frame was inadequate? Or does the new frame simply sound better?
Confident implications. Does it close with statements about what will happen or what must change? Do those implications follow from the preceding argument? Or could they have been written without the middle section?

If the opening could have led to the closing without the material in between, the middle was performance, not reasoning. The output passed through the shape of analysis without executing it.

4.1 The Comparison Test

A more rigorous version: take the same question and run it through two separate AI conversations. In the first, present it the way you normally would. In the second, strip out any indication of what you think the answer is and ask for competing interpretations rather than a single analysis.

Compare the outputs structurally. If the first is a fluent, confident argument and the second is a messier, more hedged set of competing positions, the first was performing the template. The second is closer to what genuine uncertainty looks like, and genuine uncertainty, as Guide 09 (Confidence Calibration) explained, is usually where the useful information lives.

4.2 The Familiarity Diagnostic

The Simplest Test Does this output look like every other AI analysis you have received? If it does, the probability that it has been shaped by engagement-optimised defaults rather than by the specifics of your problem is high. Genuine analysis of different problems should look different. When the outputs converge on the same rhetorical structure regardless of the question asked, you are looking at a system producing templates, not analysis.

5. What To Do When You Spot It

Recognising the template is not a reason to reject the output. The content may be correct even when the structure is default. It is, however, a reason to verify, which is a different and more demanding thing than reading it and feeling persuaded.

Ask for the reasoning chain, not the conclusion. If the AI produced a confident negation (“This is not X, it is Y”), ask: “What evidence supports the claim that the conventional view is wrong? What is the strongest case for the conventional view?” This forces the model off the template. If it cannot produce a substantive case for the view it just dismissed, the dismissal was rhetorical, not analytical.

Ask for sources on the staccato claims. Pick two or three declarative assertions and ask: “What is your source for this? How confident are you in this specific claim?” As Guide 09 (Confidence Calibration) explained, the model will not volunteer uncertainty, but it can often express it when directly asked. The claims followed by hedging are the ones filling a rhetorical slot rather than reporting evidence.

Check whether the conclusion follows. Cover the middle of the output and read only the opening and closing. If the closing still makes sense, and would make equal sense attached to completely different middle paragraphs, the argument is not load-bearing. The conclusion was predetermined, not derived.

Push toward specifics. The template thrives on abstraction. “Every institution faces this challenge” is template language. “Specifically, your compliance team’s current review process takes 14 hours per submission and the AI-assisted version takes 3, but the error rate data is not yet available” is analysis. Use Guide 05’s specificity gradient: sustained specificity kills the template because specifics require actual knowledge rather than rhetorical structure.

6. Why This Matters Beyond AI Detection

There is a version of this guide that treats the template purely as an AI detection tool: here is how you spot AI writing. That version misses the more important point.

The template works on humans for the same reason it works in AI training: we respond to the structural markers of good reasoning more readily than we respond to good reasoning itself. That was true before AI existed. Consulting reports, thought-leadership articles, conference presentations, and executive summaries have been using versions of this template for decades. Humans wrote them. Humans found them persuasive. Humans promoted the people who produced them.

What AI has done is make the template cheap. When producing the shape of insight required effort, the effort itself was a partial signal of competence. Now the shape can be generated in seconds. The proxy has collapsed. The markers remain. The thing they measured has become optional.

This connects to the broader argument of the Mirror Effect: AI does not create new problems in how we evaluate quality. It reveals where our evaluation was already depending on signals that correlated with quality rather than measured it. The writing template is the most personal and immediately observable version of that revelation. Every time you read an AI output and feel informed without being able to say specifically what you learned, the template has done its work. Noticing that feeling is the beginning of working differently.

7. Key Takeaways

The default AI writing structure is a template, not analysis. It follows a recognisable pattern: opening negation, staccato assertions, analogy pivot, confident close. Each element substitutes for a genuine reasoning operation.

The template was not invented by AI. It was already dominant in human thought-leadership and consulting output. RLHF training identified it as what humans rate highly and the models industrialised it.

Recognising the template is a verification trigger, not a rejection criterion. The content may be correct. But template-shaped output has not been tested against the question it claims to answer. Verify the claims, check the reasoning chain, and push toward specifics.

The deeper lesson is about your own evaluation habits. If the template persuades you, that tells you something about how you assess quality in general, not just in AI interactions. The same structural markers that make AI output feel insightful make human output feel insightful, and in both cases, the feeling is not evidence.