Why every statistic in our articles has a source

Generated articles have a reputation problem, and most of it comes down to numbers. A confident sentence with a percentage in it reads well and is easy to produce. Whether it is true is another matter, and the reader has no way to tell.

We decided early that InkApex would not publish a statistic it could not point at.

How it works

When the engine researches a keyword, it reads the pages that currently rank for it and pulls out the figures and claims that matter. Each one is stored with the URL it came from. When the writer model drafts the article, it may only use numbers from that set, and the draft carries the link for each one.

Facts about your own business work the same way. They are extracted from your pages, each with the page it came from, and you can switch any of them off.

What the reviewer checks

Before a draft reaches you, a second model reads it against the research and your facts. It looks for numbers with no source, claims that go beyond what the source says, and statements about your business that your site does not support. A draft that falls short goes back to the writer once, then comes to you with its score and the reviewer's notes.

What this costs

Some articles end up with fewer figures than a human writer would reach for, because the sources did not support more. We think that is the right trade. An article you can defend is worth more than one that sounds authoritative.