Generative AI and critical thinking in marketing

AI can produce a convincing answer before you've had time to question it.

By MarTech Logic

Strategy, campaigns, technology and measurement

Quick answer

AI can produce a convincing answer before you’ve had time to question it. That’s the reason to keep checking the evidence, not proof that using AI damages your thinking. A 2025 study looked at workers’ reported effort; it didn’t establish a lasting loss of ability.

What the research found

A 2025 study by Microsoft Research and Carnegie Mellon researchers surveyed 319 knowledge workers. It examined their reported critical-thinking activity and effort when using generative AI, including the relationship with confidence in the technology and in their own ability.

This was a study of what people reported doing at work, not proof that AI causes lasting damage to their thinking. The immediate editorial question is simpler: what are you letting the tool decide without checking?

Review the claim, not the confidence

Checks before publication
MaterialCheckEditorial action
StatisticOriginal source, date and denominatorRecalculate and retain the qualification
ComparisonThe same criteria for every optionRemove unsupported rankings
QuotationExact words and original contextCorrect or remove the quotation
RecommendationEvidence and plausible alternativesExplain why this choice fits the brief

A source can be genuine and still fail to support the sentence attached to it. Open the underlying report, read the relevant passage and check whether the draft has widened its claim. A survey of one customer group should not become a statement about an entire market.

Treat disagreement as part of the work. If two sources use different definitions, explain the difference rather than asking an assistant to blend the numbers into one confident answer.

Keep an independent view of the brief

Write down the audience’s question and the evidence needed before generating copy. This gives the editor something firmer than the draft’s own argument to judge it against. Ask which finding would change the recommendation and whether the source pack contains that evidence.

For campaign reporting, reproduce the important calculations from the data. For a product comparison, verify the feature in the relevant plan or documentation. For an opinion piece, make the reasoning clear enough that a reader can disagree with it.

Measure time saved after checking

Count verification and rewriting in the production time. Keep a short record of recurring errors, such as missing qualifications or unsupported comparisons, and improve the brief accordingly. A faster first draft is useful when it leads to sound published work. Our guide to AI-assisted marketing research covers the source-checking process.

Use human review where it changes the outcome: verify repurposed claims and count only work the team can actually accept. See our guides to AI content repurposing and AI marketing ROI.

MarTech Logic newsletter

Know what deserves your attention.

The shifts in marketing worth understanding. The tools worth questioning. The ideas worth taking into your next meeting.

Get occasional MarTech Logic emails from ABM Logic Ltd. Unsubscribe at any time. Privacy policy.

Recent reads