Editorial policy
Updated · By llmwise, AI-assisted.
How we write and check these pages
A model drafts each page from data we hold: our plans, model catalog, recorded test runs and other companies' own pages, quoted and dated. Automated tests and a reviewing model check every change, and the llmwise team approves it before it goes live. A page whose outside facts pass 45 days leaves search until they're rechecked.
Who writes them
llmwise, the company that makes this chat app. The guides, comparisons, price pages, data studies and launch notes are signed by the team, not by a named writer: they're the company's work, and the company answers for them.
We make llmwise, so we're one side of every page that sets it beside another app. On those pages, each fact about the other app is quoted from its maker's own page, with a link and a date.
Why say how a page is made? Google's guidance on helpful content suggests it for AI-assisted pages: “Sharing details about the processes involved can help readers and visitors better understand any unique and useful role automation may have served.” Its note on AI-generated content adds: “Appropriate use of AI or automation is not against our guidelines.”
How is a page made?
The same steps, in this order, for every page.
The numbers come from code. llmwise's prices, counts and limits are read from the plan settings billing uses, so a page can't promise a count the app doesn't give. Model prices and context windows come from our catalog, last checked October 7, 2026.
Other companies' facts are quoted, not retold. A price or a limit is copied word for word from that company's own page, with a link and the day we read it. When a fact comes from the press instead, the page says whose claim it is.
Test results come from our own runs. Every model answered the same 50 prompts on September 27, 28, 29 and October 2, 7, 8, 2026. Fixed checks score 30 of them; Claude Opus 5.5 grades the other 20 by a published rubric, and GPT-6 Astra grades Claude Opus 5.5's own replies.
A model writes the first draft from those facts.
Automated tests run on every change. A page fails them when a figure no longer matches the data, a title runs too long, or it repeats too much of a sibling page.
A model reviews each change against the data and the sources it cites. The llmwise team approves every change before it goes live.
The prompts, the scoring rules and every result are on our test runs page.
When each fact is checked again
A page whose facts about another company are more than 45 days old stays up but leaves search until they're checked again. Test results follow the same rule.
| Fact | Where it comes from | Last checked | Next check |
|---|---|---|---|
| llmwise's prices, message counts and limits | The plan settings llmwise bills from | On every change to the app | Never stale: the pages and the app read the same numbers |
| Models' API prices and context windows | Each maker's own price page, and OpenRouter's for the hosts it routes to | October 7, 2026 | When a maker changes a price, or a model is added |
| Other companies' plans, prices and limits | Each company's own pricing page, help center or announcement, quoted word for word | September 28, 2026 | November 12, 2026, or the page leaves search |
| Other chat apps' plans and features | Each app's own pages, quoted word for word | September 28, 2026 | November 12, 2026, or the page leaves search |
| Test results, costs and reply times | Our own runs: 50 prompts on each of 18 models | September 27, 28, 29 and October 2, 7, 8, 2026 | November 11, 2026, or the results stop counting |
| News (9 launch pages) | The company's own announcement, linked and dated | The day each page was published | When the company changes what it announced |
Each page's “Updated” date, by its byline, is the day its text last changed.
What we don't do
Print a number we can't trace to our own data or a dated source.
Judge an app we haven't used. Where an app is good is what its own page says; who should pick which is marked as our view.
Pretend to be neutral. We make llmwise: beside another app, the app's facts are its own words, linked and dated, and any verdict is ours.
Re-date a page to look fresh. A title carries the month its facts were checked, and “Updated” is the day its text last changed.
What's written here
The pages written this way, by the index that lists them.
About how we write
Are these pages written by AI?
Yes: a model drafts the text from our data and from other companies' pages, which we quote and date. Automated tests and a reviewing model check each change against that data and those sources, and the llmwise team approves it before it goes live.
Why isn't a writer named?
No. The pages are the company's work and llmwise answers for them, so the byline is the team's rather than one writer's.
How do I report a mistake?
Email hello@llmwise.ai with the page and what's wrong. We check it against the source, fix the page, and its “Updated” date moves with the change.
Found something wrong?
Write to hello@llmwise.ai with the page and the fact. Or try the app these pages describe: 5 messages are free.