Why a €200 brand gets recommended beside your €2,000 one

Ask a luxury brand who its competitors are. You'll get three or four names, all of them expensive.

Ask ChatGPT the same question, phrased the way a buyer phrases it, and the list comes back longer. Some of those names cost a fifth as much.

Both lists are real. Only one of them decides whether the brand gets recommended.

With our partner Redify, we take this shift and ask: once AI is doing the answering, who are you actually competing with?

The lists differ because of how buyers now ask.

The Osservatorio Search in Italy 2026, which surveyed 1,003 online adults in March, documented the shift, citing examples from interviews. A buyer who once typed "best running shoes" now supplies a full brief.

52 years old, 85kg, a road runner three times a week covering 10 to 15km. Needs stability for overpronation, good shock absorption, and high durability. Wants long-run models separated from speed-work models, with pros, cons and a price range.

Google reported at I/O 2026 that the average AI Mode query now runs 3 times as long as a traditional search query. Planning-type questions are growing 80% faster than the AI Mode average.

Look at what that brief leaves out: no market segment, no tier, no mention of premium or entry level.

Why a brief breaks the logic of a competitive set

Here's the part that makes tier irrelevant, and it's architectural rather than editorial.

Google's AI Mode doesn't answer that brief as one question. It uses what Google calls query fan-out.

The model breaks the prompt into multiple sub-queries. Those run concurrently against the live index, the Knowledge Graph and the Shopping Graph. Independent estimates put a standard decomposition at 8 to 12 sub-queries.

So the brief above becomes a set of smaller, separate retrievals:

• one for stability in overpronation
• one for shock absorption
• one for durability over high weekly mileage
• one for the long-run versus speed-work distinction

Each of those runs on its own merits and pulls its own candidate set.

A brand priced at a fifth of yours doesn't need to compete with your positioning. It needs to win one sub-query, on durability or on stability, and it enters the answer through that door. Tier never filtered anything, because tier was never one of the sub-queries.

In the AI era, brands don't compete for rankings — they compete for relevance across thousands of customer intents. Winning requires more than visibility; it requires a digital architecture that AI can consistently understand, trust, and recommend.

— Giuliano Torelli , CMO, Redify

This is also why the answer feels coherent while the sourcing is scattered. The model synthesises across the sub-query results, so the buyer reads one confident recommendation assembled from a dozen separate retrievals they never see.

But surely the budget filters them out?

This is the obvious objection, and the answer is the sharpest thing in this piece.

A stated budget filters in one direction only. Say €500 and everything above it drops out. Say €2,000 and nothing below it does.

The cheaper brand survives a premium buyer's budget constraint. The premium brand does not survive a mid-market one.

Premium brands are structurally exposed by that asymmetry in a way budget brands never are. And most briefs don't lead with price at all. They lead with durability, longevity, performance and reliability, which are precisely the grounds on which a well-made cheaper product competes credibly.

Why enterprise teams can't see it happening

A competitive set is a considered document. It gets agreed internally, reviewed annually, and built by people who know their market properly.

AI rebuilds it per question, per engine, per month, from sub-queries the marketing team never sees.

Measurement widens the blind spot rather than closing it. Share of voice is usually calculated against a competitor list defined in advance, which works well for tracking a rivalry you already understand. A brand you never listed can't lower a score it was never part of.

Then there's the engine problem, which defeats even honest measurement done on one surface.

An analysis of 730,000 response pairs found Google's AI Overviews and AI Mode cite the same URLs just 13.7% of the time. The two surfaces still reach 86% semantically similar conclusions. An audit of 680 million citations put the overlap between ChatGPT and Perplexity at 11%.

Repetition doesn't rescue it either. Running the same AI Mode query three times returns an identical URL set only 9.2% of the time. On 21.2% of queries, every attempt returns different URLs.

One engine, checked once, tells you almost nothing about the other four.

What the wider set costs

The Italian data moves in one direction as research deepens.

• Among consumers who search roughly as they always have, 9.9% bought a brand other than their habitual choice.
• Among those doing much deeper AI-assisted research, that reaches 75.3%.
• In the same group, 42.3% discovered and bought a brand they hadn't previously known.

The report's own reading is worth adopting. Loyalty erodes because the depth of the question changes, and most companies aren't prepared to answer a deeper question usefully.

In the previous model, a brand won on habit and recognition. In the generative model, the AI runs the comparison for the buyer. The brand has to show it's the better choice rather than the better known one.

The slots are finite, which is what turns this into a real loss rather than a shared shelf. AI Mode responses average around 7 unique domains in the citation sidebar, and other engines synthesise from fewer still. A handful of names get recommended and the rest of the category is absent.

This holds even where buyers keep hold of the purchase. McKinsey's January 2026 analysis of delegation looked at high-consideration categories such as luxury goods.

Buyers use agents to research, compare and analyse, then decide and transact themselves. The agent still surfaces alternatives and better price points.

The comparison happens either way. Only the checkout stays human.

Optimising for intent

1. Build content that answers sub-queries, not campaigns. If a brief decomposes into 8 to 12 retrievals, a single comprehensive page competes for one of them and loses the rest. Modular sections of 200 to 400 words under question-shaped headings give each sub-query something clean to extract.
2. Match the format to the intent. Commercial questions draw 40.86% of their citations from listicles, and in professional services, 80.9% of those citations sit on third-party properties rather than brand-owned domains. Informational questions favour long-form articles; transactional questions favour product pages. A content plan built for one stage quietly loses the others.
3. Choose the intents deliberately. For a premium brand, maximum presence is the wrong target. The model's picture of your brand is assembled from the contexts it repeatedly finds you in. Constant co-appearance with mass-market alternatives teaches that association rather than just displaying it.

The work is winning the intents that carry margin, and staying deliberately absent from the ones that don't.

How we approach it

This is why Competitive Tracking in Search Bridge works from intents rather than a fixed list of prompts. Each intent expands into multiple variations and runs repeatedly across the major AI engines. That repetition is the only way to separate a real position from run-to-run noise.

It's also why the Brand Perception Module scores Uniqueness as one of its six core KPIs. Uniqueness registers whether AI still treats a brand as distinctive or has quietly folded it into a generic set of alternatives. That folding is precisely what intent-level competition produces.

Every finding lands as a prioritised Activation, ranked by expected impact.

The question for your next competitive review

Most competitive reports list the brands a team already watches. The more useful question for a board is which brands AI puts next to yours, on the intents that matter to your margin.

We can show you that: your share of voice on your priority intents, against the competitors AI actually returns, across the major engines.

That’s how you take control of your AI representation.