Special Edition: Who It’s For, Who It’s Not For – Ernesto Ortiz’s Product Page Play for AI Recommendations

Your buyer is asking AI which dishwasher to get before they ever see your product page. That’s what Ernesto Ortiz of Structural SEO told me on a Special Edition lightning round of Unscripted SEO, his return visit after his first full interview on the show. In ten minutes he covered what’s changed: AI is now the recommendation layer, rank data is wobbling, and the product pages that get picked are the ones that say who a product is for and who it isn’t for. Here’s that play, step by step, so you can run it on your own site or a client’s. The full episode recap is on Unscripted SEO.

Why the product page is where AI recommendations get decided

Ernesto’s starting point is that people search differently now. AI isn’t only for the tech savvy anymore. His mom and dad use it to ask what to buy.

There’s a new recommendation intermediary between interest and purchase, and it’s driving people’s decision. – Ernesto Ortiz

He’s furnishing a new apartment right now, and that’s his example. He asks AI which model fits the price, goes with what it tells him, then uses Google to check availability, pricing and reviews. The AI step doesn’t send you traffic, so it’s easy to decide it doesn’t matter. Ernesto says it’s really influencing decisions. If the model is the one doing the recommending, your product page is the source material it reads.

Run the who it’s for, who it’s not for audit on your product pages

Ernesto has done this with a couple of clients, across most of their product pages. He says not a lot of people are doing it. Here’s the workflow as he described it, with the bottom-of-funnel layer I added on tape.

1. List every attribute and feature. In his words, “every attribute and feature needs to be thoroughly covered and explained” in a way that’s simple to understand.
2. Answer four questions on each page. Who’s it for? Who’s it not for? What does it include? What doesn’t it include?
3. Say who it’s not for. Do it literally if it’s clear, or by inference if it isn’t.
4. Cover the bottom of the funnel. Refund information, shipping, packaging. The details a buyer checks right before they buy.
5. Use AI to scale it. It doesn’t matter if you have a thousand products. AI makes it efficient enough that every one of them can get a detailed entry.

Here’s what step three looks like in practice. A bucket of interior paint can say it’s for homes with long winters, anybody in the northern latitudes or in Canada dealing with a lot of snow. A software product can say it’s for product managers who already use additional analytics platforms. You’ve just told a model which question your product answers.

Who’s it for? Who’s it not for? What does it include? What doesn’t it include? – Ernesto Ortiz

Ernesto didn’t share a results figure, only that “we’ve just had great results with that.” Treat it as a practitioner’s report, not a benchmark, and measure it on your own pages.

Write product content for the long tail AI can answer

The reason this works is the long tail. Ernesto says it’s getting bigger because people can now ask niche questions about their specific situation, and AI can resolve them. Ask Google for the best shampoo for a thirty-year-old with a particular problem, climate and hair type, and he says it would be a terrible result. AI gives a precise answer.

If you don’t provide very specific content in the very specific places you need to provide it in a very clear way, you’re not getting picked up. – Ernesto Ortiz

That’s how small, niche brands show up above massive industry giants. They went specific and, as Ernesto puts it, “they’ve actually defined it and talked about it.” You can’t win that query with a generic product description.

Don’t make product page decisions from rank data alone

You’ll want to measure the audit, and that’s where Ernesto gets cautious.

It seems we can’t trust tools as much as we could a couple of weeks ago. – Ernesto Ortiz

He hasn’t decided whether the poisoned results he’s seeing are a bug or intentional. What he’s sure of is that results for automated requests don’t match what a human sees, and he asks how you’re going to trust results that are thirty percent off. That number is his framing of the gap, not a measurement. His fix is going back to manual checks.

Search Console won’t fill the gap either. My dark-click study of 38 Search Console sites, covering 12 months and about 4 million clicks, found the mean site gets no query data for 63.6% of its clicks, and the median site 72%. John Mueller confirmed to a search community Slack I’m in that these are anonymized queries, ones searched fewer than a couple dozen times a month, and that the anonymization isn’t going away. The long tail Ernesto is writing for is the same long tail GSC hides. So judge the audit on what you can still see on the page: sales, leads and conversions from the products you rewrote.

Product page SEO for AI search starts with who it’s not for

So pull up your top ten product or service pages and ask the questions Ernesto asks. Does each one say who it’s for? Does any of them say who it’s not for? Would a model reading only that page know which shopper to send to it? If the answers are no, you’ve found your next sprint. If you want the bigger picture on getting named by a model, read how to get recommended by an LLM before the training data closes.

I wrote up what this conversation changed in my own thinking on jeremyriveraseo.com. You can find Ernesto at Structural SEO and on LinkedIn.