What do you do when the tool that decides whether you get recommended has never heard of you?
That was the actual question on the first Unscripted SEO panel, and it did not come from me. Ben Senescu asked it, out loud, of two people who have been doing this a lot longer than he has. He runs OpenSEO, he started in February, and he wants ChatGPT and Claude recommending his tool when somebody asks what a beginner should use. So he put it to Patrick Stox and Ben Wills: what should I be doing over the next month?
The answer is the most useful thing anybody has handed me on this show in a while, and it applies to you whether or not you build software.
You are optimizing two different surfaces on two different clocks
Patrick split the problem in half, and the split is the whole insight.
“You’ve got a while before you’re really in the training data as a known entity. For now, it’s the retrieved pages where you’ll want to be found.” — Patrick Stox
Being in the model is a slow, mostly passive outcome. Being in the pages the model goes and fetches at answer time is a thing you can work on this week. Those are two surfaces, and they run on two clocks, and most of the advice floating around right now mashes them together into one anxious to-do list.
Here is what actually falls out of that when you sit with it:
- Go find the pages that already answer your question. Not your keywords — the actual roundups, comparisons and “popular tools” lists that a model is going to pull when somebody asks. Patrick’s instruction was to make sure there is a call-out section on those pages that says, in effect, here I am and here is what makes me different.
- Lead with the thing nobody else can say. Everybody knows Ahrefs, Semrush and SE Ranking, and models will happily repeat that forever because it is common knowledge. Patrick’s line to Senescu was “you’re the only open source one.” That is not branding. That is the one attribute that makes including you add something to the answer instead of padding it.
- Stop waiting to be famous first. The training-data question is real, but it is not the one you can act on today.
The lists are the link building now
Ben Wills took the same answer and turned it into a Tuesday afternoon.
Find the “awesome lists” on GitHub — there is one for practically every category, including open source and marketing tools — and get onto every one that fits. Then find the publishers who write about free tools, or small business tools, and do one of two things. Ask to be added to a list they already have. Or offer to write the piece that puts you next to the other options, framed around helping their audience rather than around you.
His observation about why that works is the part worth stealing: being free, or open, or genuinely cheap opens doors that stay closed to a paid product. People share things they can hand to a friend without a caveat. If your only pitch requires somebody to spend money before they can evaluate it, you are asking for a much bigger favor than you think you are.
And Patrick added the mechanical version. The tools are shopping. When somebody builds something and needs data, they go looking for what is free, what is low cost, what has a free tier — and there are directories those systems check. A free tier is not a discount. It is a distribution channel with an API.
One tiny word changes which companies come back
This is Ben Wills’s own test, and he corrected me on the tape when I misattributed it, which is exactly what you want from a panel.
Tell a model you are flying to Los Angeles and renting a car, and the hotel it recommends depends on which car you named. Corolla and BMW produce different hotels from an otherwise identical question. That alone should change how you think about “ranking” in these systems.
Then it stops being about the prompt at all:
“ChatGPT saves little things to memory… you can talk four conversations ago about going to LA and buying a Civic or a Ferrari or whatever it is, and then when you ask for hotel recommendations, parts of that conversation could be used as context in a completely different chat. The personalization aspect is orders of magnitude more pronounced with LLMs than it is with search.” — Ben Wills
Search was words. It was words all the way down to how the data sits on disk and how an inverted index gets built — phrase length, specificity, match. An LLM gathers everything it can, forms a sense of what your business is, and refers back to that sense while it writes. You and I can send the same prompt and get different companies, and neither of us did anything wrong.
So when you check your visibility in one of these tools and see a number, remember whose number it is. It is one person’s, in one context, on one day.
The new spam systems might not forgive you
Patrick’s soapbox turn is the one I keep repeating to clients.
“Any tactic that is abused will be shut down… When Penguin came along it hurt for a few years, and then Google just said, all those bad links you built, we just forgive you. That may not be the case in all these new systems. They might hold a grudge for the next 10, 20, 30 years for all the spam that you did in the past.” — Patrick Stox
He called listicles as the first casualty of the new spam systems, and his tell is simple enough to use as a filter: anything that works too well is going to get knocked down a few pegs, if not fully penalized. So the question is not whether the tactic works. It is whether you can afford to still be carrying it in five years, in a system that was designed from day one to remember.
Ben Senescu named the two pitches that are landing in your inbox right now — two thousand a month to spam every Reddit forum, or a hundred a month for an AI backlink exchange — and made the honest case against them. They might work briefly. They are a distraction from learning how to actually market the thing.
Do the boring year
Ben Wills has been in this since August 2001, and his advice to anybody starting now is a kitchen metaphor.
If you are going to be a sushi chef, the first year you just make rice. You do not touch the fish. He put his analysts on keyword research and nothing else, because doing it properly across ten, twenty, fifty sites is how the relationship between a keyword and an intent stops being theory and starts being instinct.
He thinks prompts are the same apprenticeship now. Sit down and work at it: how do I get this thing to answer this way, how do I get my site into that answer, what do I have to say. That will build a better intuition for AI optimization than any framework anyone is currently selling you.
Which is a strange kind of good news, if you have been feeling behind. The entry fee is attention, not budget.
What I would actually do on Monday
- Pull the five to ten pages that would get retrieved for the question your buyer actually asks. Not your head term — the question.
- Get yourself on those pages, accurately, with the one attribute that makes you different from the names already there.
- Find the lists and directories in your category that are free to join, and join them.
- Write down, in plain words, what you do — because a model that has to guess will guess.
- Run the same prompt from three different accounts before you believe any visibility number.
None of that is a tactic that gets abused. Which, per Patrick, is exactly why it is still going to be working in ten years.
The full panel is here, with Patrick Stox, Ben Wills and Ben Senescu. Their solo episodes are worth your time too: Patrick on building in the GEO era, Ben Wills on one-word prompt testing, and Ben Senescu on open source SEO and the $99 ceiling.