Jev doesn't write anything. Still, it's a breakthrough for SEO/AEO.
You give it a question and a set of options, and it picks one, very fast and very cheap, with a confidence score.
Which is essentially what a big part of SEO work is — lots of small decisions over long lists.
Some ways to use Jev for SEO:
1. Classify keyword intent. Give it your keyword list and the options informational / commercial / comparison, and you get an intent map without burning your expensive model's tokens on it.
2. Decide the page format per keyword. Does this one want a listicle, a comparison page, a guide, or a glossary entry? That's a pick-one decision, so it fits.
3. Triage what to refresh. Feed it a page's target keyword plus its search console trend and let it sort your content into refresh / leave / consolidate, then you only look at the refresh pile.
4. Filter backlink and outreach prospects. Relevant to your niche or not — going through 500 sites by hand is exactly the kind of work this should take over.
5. Sort community threads. Pull Reddit and X threads for AEO research, and most of the filtering question is just "could you add value in this discussion or is it noise?" A cheap classifier in front means the good threads reach you and the junk doesn't.
6. QA programmatic pages before they go live. On-topic and substantial, or thin? Better to catch that with a half-second check than with a ranking drop later.
One caveat: the accuracy depends a lot on how you phrase the question and the options.
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution