okaihub

Do AI Tools Maintain a Blog or a Changelog? We Checked 20,946

45.5% of AI products run a blog and only 12.2% maintain a changelog, measured across 20,946 products — what those gaps signal to buyers.

Published 2026-09-25 · 2 min read

Two of the cheapest trust signals a software product can send are also the least common. Across 20,946 products in our directory where we could check, 9,537 run a blog (45.5%) and only 2,551 maintain a changelog (12.2%).

These are presence checks, not quality judgements — we recorded whether the page exists and is reachable, not whether anyone reads it. But existence is the part a buyer, a reviewer or an AI assistant can actually observe, and it is the part most products skip.

Why a changelog matters more than a blog

A blog tells a visitor you can write. A changelog tells them the product is alive. For a buyer evaluating an AI tool — where models, APIs and pricing shift constantly — evidence of recent maintenance is closer to the question they are actually asking, which is whether this thing will still work in six months.

It is also the rarer signal by a wide margin. At 12.2%, a maintained changelog puts a product in a small minority of its category, and it costs almost nothing: a dated list of what changed.

SignalProductsShare of 20,946
Runs a blog9,53745.5%
Maintains a changelog2,55112.2%
Presence only. A blog with one post from 2023 counts the same as one updated weekly.

What the gap says about the category

The blog figure is close to half, which suggests it has become a default expectation — a site without one looks unfinished. The changelog figure at 12.2% suggests the opposite: it is still treated as optional, or as internal documentation that never gets published.

That asymmetry is an opportunity rather than a criticism. Publishing a changelog is a few hours of work and puts a product ahead of roughly seven in eight of its peers on a signal that buyers weigh heavily. It also feeds search: each release note is an indexable page with a date on it, which is the freshness signal both Google and AI assistants look for.

If you maintain neither

  • Start with the changelog. It is smaller, more useful, and rarer. One entry per release, dated, in plain language about what changed for the user.
  • Then write only when you have something to say. Four substantive posts a year outperform forty thin ones, and a blog that stopped updating in 2024 is worse evidence than no blog at all.
  • Put both where they can be found. A changelog behind a login, or a blog three clicks deep, does not do the work — and cannot be seen by a crawler either.

More in Measuring growth