In money prompts, brand awareness stops being enough

“Gemini, what mattress would you recommend for a side sleeper with back pain?” “Tempur is the king of high-density memory foam”. Good, you think, so you’ll buy it. If you keep reading, though, you discover that “there are excellent alternatives on the market, often at a lower price”. In one line the assistant has crowned a brand and, with the other hand, pointed you towards something cheaper.

A professional in the field reads that addition and knows how to weigh it: they know what’s behind Tempur and a price comparison doesn’t sway them. Someone who turned to AI precisely because they lack that background, however, takes the advice at face value and goes looking for the alternative.

It’s the same answer, and it sends two people in opposite directions. This is where the money prompt game is played, the commercial questions asked by those about to buy who haven’t yet chosen a brand: the well-known name enters the answer, but for the less expert reader it stops acting as a guarantee. The reason has a technical root, and we’ve brought it to light with a practical analysis.

A well-known name enters the memory, but the purchase advice is sought elsewhere

A famous brand is already inside the AI. Its awareness was absorbed during training, along with millions of pages, and the model retrieves its name effortlessly whenever its category comes up.

It doesn’t need to search for it: it knows it by heart. This is true for Tempur among mattresses just as for Miele among washing machines: in every market, well-known names surface in answers earlier and more often than the rest.

The advice on what to buy, however, comes from elsewhere. When the question asks for a judgement – which is the best, which is worth it – the model knows its memory isn’t enough, since it risks repeating something outdated, and so it runs a web search while answering, hunting for external confirmation.

The brand’s fame, which got it through the door of the mention, doesn’t come into play here: the moment the assistant reads comparison sites and reviews to decide what to recommend, the well-known name has no say left. Anyone measuring their visibility by counting mentions stops at the door and thinks they’ve seen the room. What happens inside, though, is what our tools reveal, and that’s where the big names’ advantage dissolves. On Tempur, to begin with, and then on three home appliance brands, with a result that repeats itself.

On Tempur, the recommendation depends on which name is in the question

We asked ChatGPT, Gemini, Perplexity and Google AI Mode about premium sleep money prompts, and the verdict on Tempur changes depending on one specific detail: which brand appears in the request.

Ask “an alternative to Dorelan for side sleepers” and Tempur wins, because the rival named is someone else and it serves as the benchmark: Gemini crowns it “the absolute point of reference”.

Turn the request into “an alternative to Tempur for side sleepers” and the tone flips, no answer recommends it and the assistant uses that name as a springboard to push towards something cheaper. Remove every brand and just ask “the best mattress for side sleepers with back pain”, the barest money prompt, and Tempur disappears entirely.

Even when it does promote it, the AI attaches a clause. “If your budget allows and you’re after maximum pressure relief, Tempur is the absolute point of reference”, writes Gemini, and the recommendation arrives already fenced in by price. ChatGPT is even blunter: “the gold standard, but valid alternatives exist, often at a lower price”. The praise doesn’t reward Tempur, it uses it as a ruler mark to measure who beats it on cost. It’s a preference borrowed from the competitor named in the question, and it vanishes as soon as the question changes.

Miele, Bosch and Candy tested on money prompts

We expanded the analysis and built a basket of money prompts on washing machines – the questions people genuinely ask before buying, from the most reliable brand to the best choice under a certain budget – putting them to the four engines with our AI Prompt Tracker.

For each brand, the tool calculates six indicators of its weight in AI responses, from presence as a source to competitive role down to the tone with which it is mentioned. Two are enough here, and it is the gap between them that tells the story: the Mention Rate counts the responses in which the name appears in the text, and shows how well the AI knows the brand; the Recommendation Rate counts those in which the AI presents it as the recommended choice, and shows how much it is preferred when the user asks for a decision.

So three brands worlds apart in price and audience end up looking alike: they appear in every answer and almost never reach the recommendation stage, each held back by a different sticking point.

Miele and the contrast between quality and price

Miele is the most praised name in its market and almost never the one chosen. Across all the money prompts monitored, it appears in every answer from the four engines, framed as the leader, with judgements bordering on enthusiasm: “universally regarded as the single most reliable and robust brand,” is how Gemini puts it, while Google AI Mode speaks of “a safe investment for high-end appliances.”

The AI’s judgement of Miele cools as soon as the question touches on price: enthusiastic when reliability alone is at stake, lukewarm the moment the user looks for a cheaper alternative. Quality is never in question, Miele stays the leader everywhere. What weighs against it is the cost, which prompts the assistant to steer towards other products. It does so without ever dropping the praise: “Miele is universally the most reliable,” writes Gemini, and in the same answer asks whether it’s “worth spending up to 2,500 euros when there are excellent alternatives at a third of the price.”

The one recommendation, that 25%, appears only on the single prompt that asks for the best rather than the cheapest, “premium for durability and quietness.” As soon as the question stops being about money, the brand returns among the recommended choices. Miele has built its identity on quality that costs more, and the AI acknowledges this without ever defending the price: it presents it as an expense to weigh up, alongside products promising similar service for much less.

Bosch, praised as a group but never on its own

The Bosch case is trickier than Miele’s, because at first glance it looks like a success. On every prompt monitored, the brand achieves full mention, leader status and consistently positive sentiment, without a single negative judgement. And yet the recommendation rate stays at zero.

The explanation lies in how the AI phrases its praise, almost always in the plural. “The three leading brands, Bosch, LG and Samsung, all offer excellent solutions.” “The brands of the German group BSH, Bosch and Siemens, are the true benchmarks of the sector.” The assistant includes Bosch in every shortlist of the best and systematically denies it first place. The final recommendation ends up spread across three or four names together, and for anyone who needs to walk away with a single washing machine, that amounts to silence. The sentiment peak, 87.5, occurs on the prompt “alternative to Miele for a quiet one,” where the German brand solves someone else’s problem. This is the same dynamic already seen with Tempur, preference borrowed from the name mentioned in the question, here in a subtler form. For the AI, Bosch is the safe bet to always keep among the options. A comfortable but commercially inert position, because the sale is closed by whoever names one brand, not by whoever lines up four.

Candy only counts if you name it yourself

The Candy case is the most volatile of the three. The tone the AI uses shifts from one extreme to the other, and the only thing that moves it is how the question is phrased.

The highest sentiment value, 64.6, and the sole recommendation appear together on a request that names a specific product, the Rapidó. If the user already has that model in mind, the assistant confirms it and the brand makes it into the shortlist. But all it takes is a comparison question, “Candy or Beko,” “Candy or Indesit,” for sentiment to drop to 4.3 and the recommendation to vanish. The pattern mirrors Miele’s, in reverse. There, a very high sentiment drops as soon as the question touches on price; here, a low tone rises only when the question removes every rival from the field. The AI describes it as “the go-to option for maximum savings,” and confines it to that definition. Faced with another budget option, it finds no argument in its favour, and settles the choice with a “it mainly depends on your budget.” The brand exists as a recommendation only for as long as the user brings it into the conversation.

The recommendation is decided in the sources, not on the brand’s website

Miele for the price, Bosch for the crowd of equals, Candy for the comparison: three ways of saying the same thing – good, but buy something else – without ever spelling it out. And the final choice is always decided in the same place, in the sources the AI consults before answering. On Miele’s reliability, the most cited domain is Altroconsumo, appearing 208 times in the monitoring, followed by Trovaprezzi; for Bosch and Candy, the same names come up again, Qualescegliere, the big retailers, the review portals. AI Prompt Tracker lists these domains ranked by number of appearances next to each brand, showing you exactly where the judgement really takes shape: not among the competitors on the shelf, but among the portals and publications the assistant considers reliable.

A manufacturer that calls itself the best on its own website counts, for the AI, as an interested party, not as proof. A consumer association that has tested ten washing machines and ranked them offers instead the independent consensus the model looks for to justify a recommendation. Miele’s reputation, built over years in the market, thus reaches the user filtered through whoever puts its price up against the rest.

From here the work splits into two directions. The first is to become one of those voices, by publishing content that AI recognises as useful information for its answer rather than promotional showcasing; with SEOZoom’s SEO for AI tools you can measure how well a page, in the model’s eyes, answers the question you want to be found for. The second is to keep a firm grip on the domains that already carry weight, through editorial work and digital PR activity on comparison sites, vertical portals, video content and the communities that AI draws on most often.

The game is played on how AI tells your story

This data should be read for what it is. It captures a single scan, carried out within a window of a few days: AI responses change over time, and the figures collected describe how the four brands appeared at that moment, not the quality of what they produce.

The strength of the picture doesn’t come from a single figure, but from the convergence: a mattress brand and three appliance brands, worlds apart in price and history, come out of the selection with the same pattern and at the same stage. An isolated case would be an anecdote; four independent cases behaving the same way point to a rule, and repeating the scans over time is what confirms it.

The number of citations measures how well AI knows a brand, and for a famous name it’s high by definition. The sale, however, is won or lost on the phrases AI builds when a user asks for advice, and on the sites it draws them from. “Excellent but expensive”, “reliable but always compared to others”, “good only for its price range”: these are judgements that no mention count can reveal, and they shape a purchase far more than mere presence does. Reading them one by one, engine by engine, and identifying which sites are dictating them, makes it possible to act where the choice is actually formed: on the content that becomes a source, on the prompts worth monitoring, on the outlets to break into before AI forms its judgement.

It’s a way of working, more than a measuring tool.