Ajdin Perco

Director of Content, Automation/AI, & Ops @ OGM

[Ajdin Perco @AjdPe]

The learning and experimentation are constant, which is what might be hard for some people to embrace, compared to the fixed processes they're used to.

But I think the payoff is worth it, given all kinds of benefits AI can deliver.

[Ajdin Perco @AjdPe]

Many people's default reaction to AI not doing what they hoped it would is to give up and do things manually.

It's as if their existing AI skepticism is confirmed, and they can report internally that using AI doesn't make sense for a given use case.

[Ajdin Perco @AjdPe]

There isn't an ideal state in which you can bring your AI. It's a never-ending process of improving your context layer, sharpening the workflows/prompts/agents or whatever approach you are taking, and getting ever-improving outputs.

[Ajdin Perco @AjdPe]

@Nickeubanks @johnandrews Weeeee thanks for that

[Ajdin Perco @AjdPe]

@Nickeubanks @johnandrews Hey man, which CMS & tech stack you used to build it?

[Ajdin Perco @AjdPe]

Get your pages crawled, indexed, served, and ranked on page one. There is so much work just in these four tasks. Be present in the retrieval layer that everything else stands on. And then look for some new tactics that go beyond these fundamentals.

[Ajdin Perco @AjdPe]

Teams burn a quarter chasing a presence on a platform that AI tools might discard the next day. Reddit, G2 reviews, Quora, LinkedIn, forums, etc.

[Ajdin Perco @AjdPe]

The practical takeaway is that instead of doing flashy "optimize for AI answers" formatting tactics, teams should first analyze and improve their presence in relevant search indexes.

[Ajdin Perco @AjdPe]

The practical takeaway is that you shouldn't build your strategy ONLY around the weakness that's actively being patched (at least that's what frontier labs claim). Build your marketing around the context-reading capability that keeps getting stronger.

[Ajdin Perco @AjdPe]

Two facts about AI search are true at the same time, and they pull in opposite directions.

1.) Models still can't reliably answer "is this true." Feed them enough confidently written nonsense, and a lot of it gets absorbed. Spam works more often than it should.

[Ajdin Perco @AjdPe]

relationships between entities, sources, and incentives fast. As a result, modern AI systems may appear naive about facts but sophisticated about motive.

[Ajdin Perco @AjdPe]

Everything else is theater.

Judge the output. If it's useful, it's working. If it's not, no amount of dash-scrubbing will save it.

[Ajdin Perco @AjdPe]

There are countless content pieces out there that have no em dashes, all the "quietly" and "ever evolving" phrases have been deleted, and yet those pieces tell nothing, motivate nobody, and influence no revenue.

[Ajdin Perco @AjdPe]

The only criterion that actually matters is whether the content does its job. Does it help the reader solve a problem, understand something, make a decision? Does it serve the brand publishing it? That's the whole test.