Is 13% AI an actual AI product or just a marketing label?
Sh0ny
31 июля 20261 min read
In short
The line defining when a product deserves the label “AI” is becoming blurred. I’ll explain why a percentage-based definition doesn’t work and what should be used instead.
The title of a discussion on Hacker News asks the question point-blank: if AI accounts for 13% of a product’s total functionality, is that enough to call it an “AI product”?
The number itself isn’t the main point. The main point is what we’re actually measuring. The percentage of lines of code written using an LLM? The share of screen time during which the user interacts with the model? The number of API calls to the inference server? Each option yields a different answer, and none of them reflects how deeply the model is embedded in the product’s value.
The real issue isn’t the threshold—it’s honesty. If the product continues to function even when the model is removed, then AI is a feature, not a foundation. If the product becomes a hollow shell without the model, then “AI product” is a well-deserved label, regardless of the percentage.
13% is a convenient excuse for those who want to slap a trendy label on ordinary SaaS with a single chat widget. Investors and users have already learned to spot this. The next question you should ask yourself is: “If GPT-5 becomes five times more expensive tomorrow—will I cut out the AI or rewrite the architecture?” The answer will reveal more than any percentage.