An AI publication that was cagey about using AI would be a strange thing. So, plainly:
We use AI tools throughout
Research, drafting, analysis and editing. That is the real process.
We do not publish a word about “human-crafted content” while running the same tools as everyone else, and we do not think using them is something to apologise for. What matters is what happens around them.
What we verify regardless
Every factual claim is traced to a source before publication. Language models are useful for structuring, drafting and analysis. They are unreliable narrators of fact, and they are especially unreliable about recent events, current pricing, and their own capabilities — which is most of what this site covers.
So:
- Claims about pricing, model capability, rate limits, context windows, and availability are checked against the live primary source on the day of writing.
- Statistics, quotes, and study findings are traced to the original. If we cannot find the original, the claim does not run.
- Nothing is cited from memory. If an article names a source, that source was opened.
What we will never do with these tools
- Publish AI-generated statistics, quotes, sources, or study results.
- Claim hands-on testing that did not occur, or generate a “test result” that was not measured.
- Present an AI summary of someone else’s reporting as our own reporting.
- Generate articles in bulk to cover keywords. Every piece has to pass the usefulness test in our Editorial Policy, and volume that makes checking impossible is volume we will not publish.
- Invent a writer, an expert, or a byline.
What we do not claim
We do not claim a multi-person human review process, because there is not one. This is a one-person publication. The verification described above is what happens, and it is what we hold ourselves to.
If we get something wrong
We will. When it happens, we fix it and mark it — see Corrections. If you spot something, tell us; it is genuinely useful.