How AI Decides Which Brands Get Shortlisted

Caleb Pedosiuk on The Superlative Podcast: The AI Layer Deciding Your Brand’s Fate. A conversation with Edward Buchi on The Superlative Podcast. Cleaned for readability; the words are the speakers' own.

FEATURED ON | The Superlative Podcast: The AI Layer Deciding Your Brand’s Fate A Conversation with Host Edward Buchi

Edward Buchi: Alright, we are now live. Ladies and gentlemen, welcome to The Superlative Podcast. I am Edward Buchi. One guest, no edits, no post-production, just a sharp live conversation about what it really takes to reach higher levels and what comes next. So let us get into it. Caleb here, founder of 79dev, spent fourteen years making brands visible to humans. Now he says that is not enough. AI systems like ChatGPT shortlist brands per category before a human ever searches, and if you are not on that list, you have already lost the deal. There is a lot we could cover today. Caleb, a big thank you for coming on the show. Do you want to start by telling us about 79 Development?

Caleb Pedosiuk: The 79 comes from the periodic table — gold sits at atomic number 79. This is fourteen years in. Back when we started, we were trying to figure out how to describe what we do. At the most basic level, things start with a blank sheet of paper, an empty SD card, a blank screen. From there a strategy comes together. People talk about napkin business plans, but before there is anything on the napkin, there is just potential. Once those ideas start to be captured — whether it is an SD card capturing photography, or now a command key and you are speaking to a computer and building through agents — we are creating value out of potential. Development of value. 79 Development.

We started in 2012, much heavier on the creative side — brand development, very visual. A few years in we realized it does not matter how strong the brand or campaign is; if it is not seen by the right people at the right time, it is like it does not exist. So we built the other arm: the tactical side — the tags, the pixels, the tracking, the campaigns, the funnels. It seemed purely analytical until you get into it, and then the creativity comes back in through the strategy. The last couple of years we have all experienced the shift into this remarkable accelerator with AI, and those two foundational components — creative and tactical — have come together in a special way. The center of the dartboard for us now is to help brands show up in AI search. If you have figured out product-market fit, you are solving a meaningful problem, you have communicated it clearly, and you have optimized your properties, then when someone asks ChatGPT or Claude for a recommendation and you are one of the names that surfaces, it means a lot of the work at the machine layer has been done well.

Edward: This is going to be a great conversation, because I come from a design and advertising background at OCAD. We studied how to build really great creative and all the technical knowledge behind it. But it was only when I got into crypto communities and startup work that I learned the concept of marketing — the funnel. You make a lot of content, you get seen by as many people as possible, some of those views convert into signups or purchases, and then people move down the funnel. I have seen a lot of marketers help with that top-of-funnel part. And you are saying there is now an even higher part of the funnel — an agent, an AI that looks at all the content out there and feeds it to people before they ever enter the funnel. Is that a fair way to characterize where we are?

Caleb: There are a couple of important things in there. One is the Henry Ford idea: if I had asked people what they wanted, they would have said a faster horse. We are leaving the horse-and-buggy era and entering the combustion-engine era — and in the middle of it, you see horse-and-buggy thinking being applied to a combustion-engine industry. Here is the example. How many times have you seen something pop up saying, take this prompt and create three months of LinkedIn content in sixty seconds? That is a horse-and-buggy application of a combustion-engine tool. For the last decade-plus, content has been human-to-human — you create a piece for another person to scan with their eyes and process. Top of funnel awareness, mid-funnel action, bottom-funnel decision. But the arrow has shifted. Most of that massive amount of content at the top no longer needs to be processed by a human first. It is processed at the machine layer first.

So before you grind out content optimized for humans, lay a foundational machine layer that covers everything machines and agents will be looking for. Then use those tools to get crystal clear, in thorough detail, about exactly what you do — even if it is incredibly boring for a human to read. One tool anyone responsible for growth should know about: look at your website for a /llms.txt record. WordPress and others are starting to auto-generate something like this. It is a shorthand of everything the site does — who you are, what you do, your product listing — and you can go deep, answering every FAQ in detail. If you are in tourism, you reverse-engineer it from a prompt like “help me plan this vacation, here are our mobility challenges and dietary restrictions,” and you lay out everything clearly at the machine layer.

Edward: Sorry, help me backtrack — where is this metadata being stored?

Caleb: Take any URL, add /llms.txt, and hit enter. Try it on a few sites — you can do it on 79dev.com to see ours. That is one aspect of structuring your data. The important thing is that all of your information is congruent — you are not sending mixed messages — so everything from your headlines through your supporting copy is consistent across the board. To answer the funnel question: the paradigm shift for me, relatively recently, was realizing we are creating content first for the machine layer in order to serve the human layer better. If we figure out what AI systems are looking for at the agentic layer, make that available on our site in a clearly structured way, and produce content that answers what people are actually searching for in our industry, then we are building the gears beneath the brand. Whenever the machines come by to evaluate a query, it is incredibly efficient for them to turn and say, this is who fits.

A quick analogy: picture a restaurant downtown with a menu clearly posted in the window. When people walk by and glance in, they can see exactly what you offer — you have saved them from going in, finding a menu, and opening it up. Having that simple and available for the machines to read is table stakes.

Edward: It is unintuitive — people do not see this happening. The restaurant menu analogy makes sense. Where I would add to it is: it is basically a robot going downtown, checking all the restaurants on your behalf, relying on the menus in the window being as accurate as possible.

Caleb: Exactly. Think of it like a concierge at a hotel. A guest comes up and says, we would love a spot for dinner tonight, what are your thoughts? A good concierge takes pride in knowing their environment — they walk the neighborhood, they stop by the restaurants. You want to make it easy for them to know what you do, so that at some point they go deeper and recommend you. I do not want to call them out directly, but there is a major creative platform whose visual library we use regularly, and I was shocked — I was chatting with Claude about integrating it via MCP, and Claude confidently said they do not do that. It was one hundred percent wrong. I bet they had not optimized that aspect, went to their site, and sure enough they had not. They are relying on the machine to thoroughly evaluate everything they do and figure out what it means, rather than simply stating: when someone is looking for this, here is what we do, and here is why we do it best, for these six reasons. And you can explain all six at length, because it is a text file at the machine layer. It is the documentation of the business — not something a human necessarily needs to read.

Edward: There is a lot of convergence here. One of the biggest things in media right now is nano-influencers — small followings, but much closer to the viewer than a mega-influencer. And I had a guest, the grandson of Marshall McLuhan, who actually lives nearby, and he and I got into how agents are, in a sense, going to replace your friends. Not emotionally or spiritually, but in the sense that you go to friends for feedback — what is the best movie to watch, the best show right now. An agent can replace that function, because it knows so much about you and can go learn from the outside world. I see that happening.

Caleb: This is interesting, because I wondered about the semantics. Typically you would ask a neighbor or your father-in-law for a roofer. Why would I now quickly turn to a tool and say, who are the best roofers in the area for this kind of house? The term I keep coming back to is trusted advisor. If you drew an umbrella over “friend,” a trusted advisor sits underneath it — not every friend is a trusted advisor, but a trusted advisor usually falls under friendship. The longer someone uses one of these agents and builds trust, the more it becomes a source of advice — a trusted advisor — until it is not. And I do not think that rules out the human component. I feel like humanity is being nudged into a new era where human relationships carry more weight. I am optimistic about it. If we do not have to spend so much time on the technical advice — like picking roofers — what does that open up in bandwidth for the conversations that actually matter? You skip the pseudo-interview of figuring out whether you can trust a given vendor.

Edward: I want to ask about a trade-off. As AI starts looking for things on behalf of humans, do you not think something is lost — the serendipity? You would not stumble on a restaurant you end up loving, or a roofer who did amazing work, if an agent is curating everything. Do you want to comment on that?

Caleb: Two things. First, on serendipity — if we define it as this almost supernatural aspect of fate or timing, then part of me wonders how big that really is. If it is smaller than AI, then maybe AI absorbs it. But if it is bigger than AI, it will keep working outside of it. That is the first thought. The second is more practical. When you see a creator who really excels at one thing, they often could have excelled at many things — it has to do with work ethic, persistence, resourcefulness, taking two different things and pairing them into something new. That creativity feeds the machine layer too. Jump ahead six months: instead of thirteen businesses realizing they should invest in showing up in AI search, you have a hundred out of a hundred wanting to be one of the top three restaurants mentioned when someone says, I am downtown, where should I go for lunch? Now the machine layer has to decide who deserves to be there, if it is not just a paid placement. That is where the creativity comes back.

So we do a foundational AEO and GEO engagement — right now it is around a thousand dollars, though that will likely go up — and then structured retainer work for clients who want to own their space. That restaurant might decide to engage at the human layer too: live events, their own content channel, the founders constantly talking with musicians and creators and builders in the city, hosting a live podcast on their rooftop patio. All of that, when it is optimized for the machine layer as well, factors back into their root structure — rather than being a gear that spins outside of it.

Edward: So as a business, if we create more interesting, more creative content, you are saying it might tip the scales for the machine layer to notice it?

Caleb: There are probably both extremes. On one end, content that is just flat — straight from AI, or from a human, but lacking soul and substance — is noise that adds to noise. On the other end is the almost black-hat approach: so optimized it is trying to hack the algorithm, and even if it wins short-term, you eventually realize that restaurant is cutting corners elsewhere too. There is a middle area where, if the heart and soul of the content is bringing something genuine — it can be lo-fi, it does not have to be highly produced — and it is in alignment with what the brand is about, then over time it compounds and positions you in a much greater way. Say it is a restaurant. The people they talk to and the nature of those conversations might be about the local food system, the local food scene, engaging the farmers. That niche works well for an audience that cares about it. But if their target audience is the tech community — it is Tech Week in Toronto right now — maybe they want to be the spot where people do business lunches. They understand their ICP so well: these are people who want to be fit, healthy, building meaningful things. So they build content around those topics, maybe host a live event where people talk about what they are building, and then make sure transcripts or article versions of those conversations connect back to the rest of their brand’s engine. Like a mycelium network or a root structure — you shout people out, link to their projects, and the digital layer sees this bistro as a genuine source.

Edward: So you do content with your farmers and suppliers — but those are partners within an already-established network. You are saying also have content around who you want your customers to be. That new Venn diagram. And you are talking about situating that content in their ecosystems, their social platforms. That is really insightful.

Caleb: One more example. The Tech Week angle works for a brand that wants to be the lunch spot during Tech Week. Years ago, my first job out of school, I booked the film commissioners’ marketing meetings out in Los Angeles and New York, and I was always looking for good meeting spots — a brand that positions for that would win there. But another restaurant might say, I do not want that at all, I want to talk to the local farmers in Ontario, know who is producing what in season, the struggles, the challenges — position as the pinnacle of true farm-to-table. Very different niche. And that is what is beautiful about this: someone coming to the city who cares about those things, or who is just curious and asks, what would be a really unique dining experience that only Toronto has to offer, and then this place surfaces — because the agent knows those things about them too.

Edward: That gets me excited, because it makes me feel like people who deeply care about others and want to build business that is harmonious with others — we are entering an era with so much potential for that. Help me circle back to the agents. You are saying: describe your business as completely as possible for a machine to read, great; have content do that too, great; but then also have the creative content that Venn-diagrams with the ICP. So that agent working for a tech entrepreneur will find that restaurant and recommend it.

Caleb: The llms.txt piece is more foundational — you set it and update it periodically. But your website, the home base, is where you go deep on the topics that matter to your ICP. Here is a client example. They build really cool structures — they will take the third floor of a condominium and build out a beautiful play structure out of real wood, a great spot for people to sit and eat. They are called Fern Kids, a manufacturer in Ontario, really great stuff. We realized we could take the procurement process and reverse-engineer the criteria a vendor needs to meet to be invited to a tender. So why not answer those questions in advance — make it abundantly clear we meet every criterion? Because at whatever level the agents are already evaluating things, when someone says, take this tender and find the top ten across Canada to invite, then narrow it to the realistic top three — we want it to be so clear that we are the answer. Here are the reasons we meet the criteria, here is why we are best, here is proof, and all the testimonials are current and well presented. So at a glance, something comes to the site and goes, this is one of them. It is like a guest walking into the restaurant and saying, I have a group of fifty, gluten is an issue, and can you also do a wedding ninety minutes away — and everything is already answered. The content strategy is fun because we are playing in the future. We do not know if the purchase happens three months or nine months from now, but we know architects and interior designers will be looking for this, based on the historical purchase pattern, so we prepare all of it in advance.

Edward: So you are thinking about content as anticipating the friction people will have in the future, and having it already sorted. We are coming to time, but I want to ask one final question. In my research, you said the stakes are high because agents will only shortlist three to five brands. Paint the picture — why is this such a huge deal for people to act on?

Caleb: Let me ask you this. When you ask an AI for something, do you like it when Claude or ChatGPT gives you a list of ten options, or do you prefer two or three?

Edward: The fewer the better.

Caleb: Right. There are times I will say, give me ten options, when I want to zoom out and I am not sure. But most of the time I love it when it is decisive. Last night I was looking for UV-protection clothing for the summer. I do not love polyester — I love merino wool, it feels great. I asked, and it came back strong for polyester: dries well, very popular. So I asked, from a health standpoint, is polyester great for your body? It said, well, what do you mean by health — UV protection, or the fabric on your body? For UV, polyester is strong, but merino wool is excellent, and with the right knit you can get a UV rating around 55. So now I am asking, what do you have for light-fitting hoodie options in merino wool? I did not want ten options. It came back with one — hands down, this one, great reviews, reasonably priced. My confidence level was already there. I grabbed the link and put it with the other things I am prepping for summer. Then I asked about another brand I like, and it said that one is pretty good as well, and we went through it. But Edward, I would look at that and ask: why did it so confidently recommend that first one? That is the point worth sitting with — why did that brand deserve to be there? Because there are probably multiple other brands sitting back going, what about us, our price is lower, our thread count is higher. They wanted to be at that point too.

That is where it is like getting fit. You have to eat well, work out, do it when you do not feel like it, do the reps. The foundational layer is getting the junk out of your diet — if you are doing things that are actively harmful, stop. That is table stakes. Then build the foundation, find exercises you would actually keep doing, find a spot you can do them, and enjoy it. Now you are creating content — bulletins, breakdowns, articles, thought leadership. You can have a lot of fun with it. That is the era we are in.

Edward: So the kind of content you can make is almost boundless, and I think I am getting it now — people demand confidence from their AI when they ask for recommendations. The game for brands is to help that AI be as confident as possible in choosing you. That has been really eye-opening, and I really appreciate all the insight you have shared, for the folks listening today and in the future. Caleb, how can people reach you to learn more about this and get their business found with AI?

Caleb: Our website is 79dev.com, and you can find me there. My email is caleb@79dev.com if you want to shoot me an email. I am on LinkedIn as well. Send me a note, let me know what is going on, and I am happy to offer some suggestions.

Edward: Beautiful. Caleb, a really big thank you for coming on the podcast. I genuinely appreciated you exposing how this all works. I never appreciated how intense marketing can be, and now with AI changing the game — I would not have learned it without you coming on this call. So thank you for that. And for the rest of the crew watching, thank you for joining. This has been The Superlative Podcast, and until next time.