Ask an AI tool which brand makes the best version of what you sell, and it names whichever page reads as clearly about that product, not the best-written one. If yours only mentions it in passing, you lose to a competitor whose page doesn't.
The concept behind that is entity salience, one of the biggest levers in SEO and AI search.
What is entity salience?
An entity is any specific thing a page talks about: a product, a brand, a material, a place, or even a concept like “sustainable packaging.”
Entity salience is a score from 0 to 1 that measures how central one of those things is to the whole page, not just how often the word shows up on it.
Why does it matter?
Entity salience matters because keyword density stopped being the main ranking signal a long time ago. And the same logic applies even harder to AI answer engines. You have probably heard “write naturally, don’t stuff keywords” a hundred times.
A page that mentions “hiking boots” forty times but never actually says anything specific about hiking boots scores low. A page that opens with a clear statement, something like “this is a waterproof hiking boot built for wet trail conditions,” and stays on that thread the rest of the way through, scores high.
How search engines and AI tools read your page
Google reads text with natural language processing, a set of techniques for pulling entities out of a sentence the way a person would, by reading context rather than just matching spelling. Google’s own Cloud Natural Language API has exposed a salience score for each entity it detects since around 2016.
The newest version of that API, a PaLM-based upgrade rolled out more recently, actually dropped the salience field from its response. Google still computes something close to it internally to rank pages.
Google’s Cloud Natural Language API demo
It just stopped handing you the exact number through that particular door. Which means the free tools people used to check “their” salience score are getting less reliable, not more. That is a good reason to learn the underlying structure yourself instead of leaning on one tool that might disappear, and it is exactly what the rest of this piece walks through.
We saw the same pattern at scale.
Across a study of 1,000 Shopify stores, brands were cited or recommended in AI answers only 9.5% of the time, and just 2.8% of all sources cited were the brand's own product page.
The rest of the credit went to publishers and marketplaces syndicating the brand's own words.
A page with weak entity salience gives an AI model nothing to point to as the origin. So the citation defaults to whoever wrote about the entity more clearly, even if that's a reviewer, not the brand.
A real example
Content at Scale, a AI content company noticed a page that used to rank well for “AI detector” had lost tens of thousands of monthly clicks. Standard fixes, better alt text, shorter copy, new backlink anchors, did nothing.
What actually moved the needle was small edits that tied the page more directly to their own brand: changing a subheading from “AI Detector Introduction” to “Introduction to Our AI Detector,” swapping “the AI Detector” for “our AI Detector” throughout.
Within about eight weeks the page had not just recovered, it added roughly 150,000 organic visits a month on top of where it started. Nothing about the underlying product changed. The page just got clearer about what it was centrally about.
When someone asks an AI tool a buying question, it breaks that question into concepts (the product, the use case, a specific attribute) and looks for pages that cover those concepts clearly and together.
Example of how an AI overview breaks down a product search based on it's specs and suggests products.
The four components of entity salience
Strip away the math and entity salience comes down to four things you can control on any product or collection page.
-
Position. Entities named in your title, H1, or first sentence carry more weight than ones buried in paragraph six. If “hiking boots” is your main entity, say it in the first line, not just the meta description.
-
Frequency.How often the entity, and clear references back to it, show up across the page. “It,” “the shoe,” and “our product” all count toward the same entity as long as it is obvious what they point to.
-
Grammatical role. Is the entity the subject of your sentences, or not? “This boot handles wet granite” puts the boot in the driver’s seat. “Wet granite can be handled by this boot” buries it. Small rewrite, real difference.
-
Spread. Does the entity recur through the whole page, or show up once in the intro and disappear? A product description that opens strong and then wanders into unrelated brand history dilutes its own main entity.
Simple diagram showing the four signals, position, frequency, grammatical role, spread, as four labeled boxes feeding into one salience score.
Weight these roughly in that order, position first and spread last, and you will land close to how Google’s models actually score a page.
Shopify stores are particularly sensible to entity salience
A blog can wander a little and still rank on the strength of a hundred long articles covering a topic from every angle.
A Shopify store on the other hand lives or dies on a much smaller number of product and collection pages, and each one needs to own a specific thing clearly, on its own, with no other page to lean on.
Entity salience in a shopify store what ties a specific product to its category and to what the shopper is actually trying to buy.
A collection page that never clearly states what it is a collection of, leaning on filters and thumbnails to do the talking instead, gives Google and AI tools almost nothing to work with.
We measured this directly as well
Across 1,000 Shopify stores, the average AI search readiness score was 42 out of 100, and category pages scored lowest of any template at 35.
Most collection pages have this problem. Largely because most show only a product grid with no text explaining what the category actually is.
A simple way to check where your own pages stand
You do not need Google’s API key or a paid tool to get a decent read on this. Open a product page and ask three questions.
-
Is your main entity in the title, the H1, and the first sentence? If not, that is the first fix, and usually the fastest one.
-
Read the page out loud. Is your product the subject of most sentences, or the object? Rewrite the passive ones.
-
Does anything in the middle or end of the page wander into a different topic entirely, a company history paragraph, an unrelated blog-style tangent? Trim it or move it somewhere else.
If you want a rough numeric estimate rather than a gut check, weigh those same four signals (position, frequency, role, spread) against your busiest competitor’s page for the same product and see where yours falls short. That is a cruder method than Google’s own scoring, but it points you at the same fixes.
Run a entity salience SEO audit with this Claude Skill
Checking every product page by hand does not scale past a handful of listings, so we built a Claude skill that runs this exact audit for you.
- Point it at a product or collection page and it scores every entity using the same four signals covered above, position, frequency, grammatical role, and spread, then tells you in plain language what the page is centrally about.
It flags entities worth adding, entities that are present but underweighted, and anything competing for attention with the entity you actually want to rank for.
- Point it at a whole product category instead of one page, and it will turn the results into a pillar-and-cluster content plan with an internal linking map.
A ranked table of entities from Claude with type, estimated salience score, and band.
Download the Claude skill here.
Say the facts plainly, not just the entities
There is a second layer to this that most SEO advice skips: how clearly your page states actual facts, not just which entities it mentions.
Google’s Knowledge Graph, and every AI model built on similar systems, pull facts from a page as tidy subject-predicate-object statements: an entity, an action or relationship, and what it connects to.
“This outsole grips wet granite” is one clean fact a machine can lift straight off the page. “Grip on wet surfaces is something this design takes care of” says almost the same thing, but buried in a passive clause, it is much harder to extract cleanly.
You do not need to write like a robot to do this well.
You need one or two sentences per section that state the core fact plainly, subject first, and then let the rest of the paragraph explain, qualify, or add color.
A vague heading like “Features” tells a machine nothing. “This boot’s outsole grips wet granite” is a complete, extractable fact on its own, and it still reads like something a person would actually say.
A quick entity salience checklist
-
Your main entity (the product, the concept, the category) appears in the title, H1, and first sentence.
-
Related entities show up nearby. A hiking boot page should also mention terrain, weather conditions, or trail type, not just the boot itself.
-
At least one sentence per section states a fact plainly, subject first.
-
Nothing in the middle of the page drifts into an unrelated topic.
-
The page has been updated recently. AI search tools lean toward pages that show signs of active maintenance, not stale copy from years back.
-
Schema markup (Product, FAQ, BreadcrumbList) backs up what the text already says, so machines do not have to guess.
None of this replaces good writing or an accurate product description. It just gives Google, and the growing list of AI tools sitting between you and the shopper, a much clearer signal about what your page is actually about. Get that signal right, and the rest of your SEO work has something solid to build on.
Not sure where your store stands?
Shero Commerce runs entity and schema audits across full Shopify catalogs and can show you exactly which pages are losing ground to competitors on this front.
Book a free consultation to find out where you stand.
