The rival was not in the peer set

Meta description: The competitor AI recommends is the one the public world has a clear reason to remember — not the better service. More on his work: Miklos Roth online.

The competitive review is a polished ritual. Three peers, tracked across quarters: their moves, their positioning, their AI announcements, summarized and compared. The deck is accurate and the rivalry is real — and in the companies I am describing, the buyer’s deciding question, asked in an AI engine before any vendor search begins, is being answered by a source that appears nowhere in the deck. Not a peer. Not a company the sales team has ever lost a deal to. Sometimes not a company at all. The race the leadership team is watching is real. It is happening adjacent to the race that matters.

A company hears that a customer asked an AI assistant for a recommendation and was told to use a competitor. The instinct is to ask what the competitor did to the algorithm. The more useful question is what the competitor did to the public record: what is visible, consistent, and distinctive about them in the places the answer layer reads.

How was the peer set inherited?

Peer sets are historical artifacts. They formed around the deals of a decade ago: the companies that showed up in the same tenders, the names in the same analyst quadrant, the firms the founders came from. The set is then maintained by the tools — the tracker monitors whoever the tracker was pointed at, and the quarterly review inherits the tracker’s list. I have rarely seen a peer set rebuilt from the buyer’s side. It is rebuilt from the market’s side: who looks like us, who is our size, who the board has heard of. The set is a mirror before it is a map.

Who can own the buyer’s question without looking like a peer?

The rival types companies miss are the ones that are not better at the service but are easier to describe. Brand familiarity matters. So does a consistent footprint of external discussion — publications, professional communities, third-party references that say clearly what the business does and for whom. If a competitor has made itself simple to remember and simple to verify, the answer layer has material to work with. A better service with an unclear public story gives it nothing.

What I can describe is the strangeness of the sources I find owning these questions. A professional association’s guidance page. A two-person specialist publication. A tool vendor from an adjacent category, educating buyers into a frame that excludes the incumbent’s model. A university lab’s explainer. None of these will ever appear in a deal review, because none of them signs the contract. They do something earlier and arguably more valuable: they define the question, the criteria, and the shortlist logic. The peers compete for the deal. These sources compete for the frame, and the frame decides which deals exist.

Why does expanding the set often add the wrong names?

The false explanation is coverage: widen the tracking, add more competitors, monitor the long tail. I have watched peer sets double and the blind spot persist, because the expansion logic is the same as the original selection logic — companies that look like us, at smaller sizes. The expansion adds more mirrors. The source that owns the buyer’s question does not look like the company in any dimension the selection logic checks, so it fails every expansion screen. The set grows more crowded and no less wrong. Watching more of the wrong race is still watching the wrong race.

What pattern marks a distorted race?

The marker I use is the anomaly in the win-loss data. In distorted races, losses increasingly end with “no decision” or with a winner the sales team barely met — a competitor who appeared late, already favored, with criteria the company did not help write. Sales reads this as execution failure: we were outflanked, outsold. I read it as frame loss: the evaluation was decided upstream, in materials and answers the company never saw, by sources it does not track. The deal was lost before the race the company was watching had begun.

My observations about why one brand appears and another does not are observations, not proof of an algorithm’s inner workings. Recommendation systems, retrieval sources, model changes, and query wording may all contribute. What I can inspect is the observable footprint: how complete the rival’s site is, which pages are discoverable, where the brand is mentioned, what unusual details make it a distinct answer. My objective is not to imitate every move; it is to understand what stable players have made visible that the client has not.

[AWAITING MIKLOS — scene: second scene]

Miklos Roth

What question rebuilds the set from the buyer’s surface?

Competitive strategy conversations start from the company: who threatens us? The question that rebuilds the set starts from the buyer: who answers the questions our buyers ask before they know we exist? The answer list is always surprising — I have never seen it match the peer deck — and building the set from that list reorganizes what the company watches, fears, and learns from. The race stops being a mirror.

The question the complaint should become: has this company supplied the public world with a clear reason to remember it — something specific enough that a person, or a system, could repeat it accurately?

PR excerpt (80–120 words, Miklos’s own words): A company hears that a customer asked an AI assistant for a recommendation and was told to use a competitor. The instinct is to ask what the competitor did to the algorithm; the more useful question is what the competitor did to the public record. The rivals companies miss are often not better at the service — they are easier to describe. Brand familiarity matters, and so does a consistent footprint of external discussion that says clearly what the business does and for whom.