AI-generated texts do not increase conversion: insights based on 1.5 million cases

Massive AI-generated texts make marketing materials smooth, literate, and easy to read, but in practice they absolutely do not guarantee growth of target actions. If your content marketing and copywriting have fully shifted to automatic generation, you risk encountering stagnation of metrics and dissolving into the uniform informational noise of competitors.

What the study of 1.5 million cases showed

Researchers thoroughly evaluated the integration of the built-in generative assistant on the largest international petition platform Change.org. The scientific work was published in the authoritative journal Nature Human Behaviour. To determine the real impact of the algorithms, the authors studied a dataset of 1.5 million appeals using the difference-in-differences method.

The results were significant: the «Write with AI» tool substantially altered the lexical composition and syntactic structure of the materials, but had absolutely no impact on final effectiveness. The neural network triggered a stylistic homogenization effect — texts became uniform and sterile.

At the same time, audience conversion to actual signatures remained at the baseline level. A comparative analysis of authors who created texts strictly before and immediately after the tool’s introduction fully confirmed the obtained conclusions.

How reliable are these data: methodology and context

In this case, a full-fledged natural quasi-experiment based on a huge sample is presented, not an isolated laboratory test of focus groups. The applied econometric approach allowed researchers to establish a direct causal link, completely eliminating seasonality and global information trends.

Nevertheless, specialists should consider the specificity of the environment. A classic petition relies on strong personal pain, social unity, and the author’s personal drama.

In the segments of lead generation or e-commerce, external smoothness of formulations solves its local tasks. However, the fundamental psychology of perception remains unchanged: mere compliance with stylistic norms cannot push the consumer toward the target action.

What this means for marketing and advertising

The implementation of typical generated texts carries hidden risks for commercial metrics:

• Reduced emotional response: artificial intelligence systematically selects neutral, averaged speech patterns. Human intonation disappears from communication, consequently decreasing consumer trust and audience engagement.
• Banner blindness to templates: total uniformity of messages erases the uniqueness of the offer. Potential buyers’ attention mechanically ignores standard neural network constructions.
• Illusion of high productivity: the speed of content production increases manifold, yet key metrics such as CTR and final return on investment (ROI) show stagnation.

Practical application: how to adapt generated content

Algorithms excel at handling rough concepts, systematizing theses, and preparing plans. However, semantic accents and emotional depth must be set by the editor.

To transform AI text into an effective selling tool, use the following step-by-step algorithm:

1. Eliminate bureaucratic language and corporate clichés. Mercilessly cut out clichés: «innovative product», «dynamically developing service», «key step forward».
2. Introduce personal experience and texture. Replace general reasoning with sensory language, measurable numbers, and realistic customer cases.
3. Activate social proof. Replace overused calls with demonstrations of a live professional community and real reviews.

AI draft: Our initiative aims to improve the park’s ecological situation through collective effort of concerned citizens.

Editorial edit: Every weekend, the children’s playground in our park is drowning in plastic cups. Let’s together achieve the installation of five bins with separate collection.

AI offer: Use an innovative CRM system to optimize your key business processes.

Offer with a pain point: Stop losing leads in notebooks. Set up automatic reminders for managers 15 minutes in advance.

Frequently asked questions

Should we completely abandon generative AI in copywriting?

There will be no benefit from a complete refusal. AI for business excellently handles brainstorming tasks, source analysis, and initial grouping of ideas. The problem lies solely in publishing raw material without authorial editing.

Why can a perfectly written text lose in effectiveness?

The psyche perceives polished syntactic constructions as background advertising noise. Live roughness of speech, professional audience slang, and direct appeal to triggers evoke a significantly deeper emotional response.

Conclusion

Grammatically flawless material without a deep understanding of consumer motivation invariably loses to precise triggers and lively delivery. Automate routine, but keep focus on your audience’s pains.

Want to know which semantic accents motivate your customers to make purchases? Analyze the tone of communication and test hypotheses for conversion growth using the professional tools of the RGB-Аналитика service.