Three Words Beat a Bigger Budget

On “running shoes for flat feet,” half the paid slots belong to a brand nobody would name in a list of famous running shoes, and the reason is printed in the product titles.

The page a buyer actually sees

Type “running shoes for flat feet” into Google. There are six Sponsored Products slots. Three of them belong to Orthofeet:

• Best Running Sneakers For Flat Feet, $139
• Best Running Shoes For Flat Feet Men’s, $139
• Best Shoes For Flat Feet Pain Relief, $119.99

A fourth belongs to KURU Footwear: Best Running Shoes for Plantar Fasciitis, $189.

Count the language rather than the brands and the pattern is sharper still. Four of the six titles describe a foot condition. The remaining two are a brand and a model.

Orthofeet is not one of the famous running brands. It holds half the paid block on that query anyway, and the match is visible from the page itself. The buyer typed “flat feet.” The titles say “flat feet.”

Put the two side by side and there is nothing left to explain.

Buyer: running shoes for flat feet. Title: Best Running Shoes For Flat Feet Men’s.

That is the whole move. There is no second half.

Now look at the page your product is on

You sell the right product. Someone is on Google describing it in plain words, right now, and nobody has answered them in those words yet.

Take “office chair for lower back pain.” Six paid product titles on that page. They carry Active Lumbar. Ergonomic. Massage. Not one of them says “back pain.”

If you make a chair built for exactly that person, that page has an opening in the only words they used, and nobody has taken it.

This is not a mistake you made. It is how search works now. The buyer describes a symptom, because a symptom is what they have. The industry writes the clinical term, because that is what the industry calls it. Nothing in the retrieval layer translates one into the other. There is no facet for “back pain” waiting to be switched on, and no algorithm update coming that turns a symptom into a specification.

And of course your listing says “Ergonomic.” You had to describe the chair, and “Ergonomic” is the accurate trade word for what you built. “Active Lumbar” is precise. The title is not badly written. It is written in the wrong language for the person trying to find it. Two vocabularies, both correct, zero overlap.

Buyers search in one language. Products are listed in another.

Bikes make it starker

Query: “best bike for climbing hills.”

Not one product title on that page contains the word “climbing.” The listings read Specialized Crux Expert. Terrel CF. Pinarello X1 105. S-Works Tarmac. Brand, model, groupset, material.

The buyer said “climbing hills.” The overlap is zero. Somewhere on that page is the right product for that rider. Not one title says so in a word the rider used, which means the word is still unclaimed.

One category already solved this

“Dog food for sensitive stomach” behaves nothing like the office chair page, and the difference is not that search tried harder there.

Look at what is on the page. Purina Pro Plan Sensitive Skin & Stomach. Freshpet Sensitive Stomach & Skin. Pure Balance Pro+ Sensitive Skin and Stomach. VICTOR Sensitive Skin & Stomach. Wholesomes Sensitive Skin & Stomach.

The manufacturers made the buyer’s phrase the product name. The AI Overview’s top recommendation on that page is purchasable right there.

Nobody waited for a taxonomy to add a filter. The category solved its own findability by naming products the way buyers ask for them.

That is the same move Orthofeet made in footwear. One category did it with an entire shelf. One merchant did it with half the paid block.

The part that gets oversold, so read it before you act

Winning on vocabulary is not the same as being the recommended product. These are two separate problems, and the flat feet page shows both at once.

On that same page, the AI Overview recommends ASICS Gel-Kayano, Brooks Adrenaline GTS, and HOKA Arahi. None of those three appear in the paid slots.

So Orthofeet won the language and is on the page, while the products the AI Overview actually names remain unpurchasable there. A merchant can win the vocabulary game completely and still not be the answer being recommended.

Fix the vocabulary because it is the half you control this afternoon. Do not assume it settles the other half.

What to do today

Open your product titles. Then find the words your buyers actually use for the thing those products solve. Not the category name, not the material, not the trade term. The phrase a person says out loud when they describe the problem: flat feet, sensitive stomach, back pain, climbing hills.

Then check whether the two lists share any words at all.

Where they do, you are findable. Where they do not, you are the office chair page: accurate, in stock, and the only sentence the buyer wrote is still open.

Orthofeet won half that paid block with three words in the product title, “for flat feet.” What a bid was behind it, nobody outside the account can see. The words are the part that is visible on the page, and the part you can change this afternoon. Knowing which three is the work. The buyer language exists, it is written down, and it is measurable. Open the graph and see the intersection: the words buyers use, resolved against the products you already have in stock.

Methodology

The four queries described here are drawn from a 20-query pilot spanning three query types, captured 2026-08-04 from live Google search results, query by query. No SERP vendor was used. Prices and titles are recorded as they appeared in the Sponsored Products modules on that date. Buyer-vocabulary presence is a binary check: does the buyer’s own head term appear anywhere in the product titles sold on the page?

Conditions, stated plainly. All captures used one browser profile on a United States connection, signed in to a personal Google account, so results carry whatever personalization that account attracts. A shopping browser extension was active and injected product panels into several pages; those injected modules are not Google results and were excluded from every count here. Placement on a page reflects an ad auction this study cannot see, so nothing here should be read as a claim about anyone’s bid or budget.

This is a dated snapshot, not a permanent property of search. Pages change, and the placement of products inside AI Overviews is changing quickly on exactly these queries. Anyone can reproduce it by retyping the query.