Skip to content
Building materials

How material grades get compared in AI answers

A specification inside a datasheet is a specification an engine cannot quote. That is usually why a retailer page is named instead of yours.

6 min read · Updated 2026-09-30

A material question almost always begins as a grade question. Which cement grade for a roof slab, which block for a wall, which primer for new plaster. The assistant answers the grade question from a standard, which it can read, and then has to decide which brand to name. It names the brands whose grade, conformity and application are readable text on a page. A brand whose specification lives inside a downloadable datasheet is absent from that sentence, and a retailer who retyped the same numbers is quoted instead.

The grade decides the answer, not the brand

In most categories a buyer arrives with a brand in mind and asks whether it is any good. Building materials invert that. The specifier arrives with an application and asks what the application needs, and the brand question comes second, inside the same answer.

This is good news and bad news in one sentence. The good news is that the first half of the answer is settled by a published standard, so an engine can answer it confidently and will look for brands to attach to it. The bad news is that the attachment step is a filter, and the filter is not quality. It is readability.

So two brands with the same grade and the same certification can get different treatment in the same answer, and the difference is which of them stated it as text.

Who is actually asking?

Three readers, in almost identical words. A homeowner building a house, the contractor doing the work and the architect specifying it all type something close to which cement grade is best for a residential roof slab. That is why this page uses specifier for all three: it is the one word for whoever decides which brand goes into the build.

They diverge on one axis only, which is price. A homeowner asks the current price per bag. The trade asks for a price list and a dealer. The same page can serve both, but only if it publishes the unit price and the dealer network rather than choosing one.

They converge on something more useful. All three want to know whether a product is sound for a specific job, and none of them can tell from a brand name. They are looking for the grade, the standard and the evidence, in that order.

The datasheet problem

The specification for most building products exists, is accurate, and is unreadable. It sits inside a datasheet, behind a download, often as a scanned table. Everything an answer needs is in there: grade, conformity, application, coverage, cure time.

An engine cannot quote it. What it can quote is the retailer product page where somebody retyped the grade into a bullet list, or the trade catalogue that summarised it, or the forum thread where a contractor stated it from memory. Those pages become the source of record for your own specification, and you have no editorial control over any of them.

The remedy is unglamorous and it is the single highest-value change on this list. Put the specification on the page as text. Keep the datasheet for the people who need the full document, and stop making it the only place the numbers exist.

A grade with no application answers nothing

The second most common failure is a page that states a grade and stops. The grade is correct. The conformity mark is there. And the question asked was which grade for a roof slab, which the page does not answer.

An assistant reading that page has a fact it cannot use. An assistant reading a competitor's page, where the same grade is stated alongside the applications it suits, has a fact that answers the question directly. It names the competitor, not because the product is better but because the sentence was usable.

Stating the application is also the cheapest form of honesty in this category. A grade that is appropriate for plastering and not for a structural slab should say so, and a brand that says so is the one an engine trusts on the next question.

Why the price in the answer is not your price

Material prices move faster than a crawl, and they vary by region. An assistant asked for the current price per bag is reaching for any figure it can find, and the figures it finds most easily are on retailer and marketplace pages.

A price on your page with no date and no region attached does not help, because it cannot be checked. A careful engine hedges it or prefers the retailer's number, which is how a figure you never set becomes your price in the answer. A price carrying the region and the date it applies to can be quoted, and it can also be seen to be stale, which is the honest trade.

The dealer half matters just as much and is more often missing. A specifier who has chosen your product and cannot be told where to buy it is a decided sale handed to whichever brand published its dealer network.

What to publish, and what to watch

Five blocks carry a material page: the grade against the standard it meets; what that grade is for; the certifications and test reports behind the claim; the price with its region and date; and where to buy it nearby. All five are things you already know and most of them are already written down somewhere less useful.

Then read the answers. Run the grade questions and the brand questions separately, because they fail differently. A grade question that never names you is a readability problem. A brand question that names you without a grade is a completeness problem.

The comparison prompts are worth watching most closely. A question like aac blocks vs red bricks for house walls is where a category is decided rather than a brand, and a brand that has published a straight comparison against the usual alternative is the one that gets to be in it.

In Ansyra

See which page is speaking for your product

Ansyra runs the grade and brand prompts specifiers actually type across the AI engines, records whether your product is absent, mentioned, cited, compared or recommended, and names the sources each answer drew on, so a retailer page standing in for your specification becomes visible.

Straight answers

Frequently asked

Why do AI answers recommend a grade but never name our brand?
Because the grade came from a published standard the engine can read, and your brand's grade did not. If your specification lives inside a downloadable datasheet, there is nothing quotable on the page, so the answer attaches whichever brands stated their grade and application as text. Publishing the specification on the page is what puts you in that sentence.
Is it worth publishing a comparison against a competing material?
Yes, and it is one of the few places a brand can create the answer rather than wait for it. Comparison prompts such as aac blocks vs red bricks decide a category before a brand is chosen. A page that states the comparison plainly, including where the alternative is the better choice, is the page an engine reaches for.
Should we publish prices if they change every month?
Publish them with the region and the date they apply to. A dated regional price can be quoted and can also be seen to be stale, which is the honest trade. A bare price cannot be checked, so a careful engine hedges it or takes a retailer's figure instead, and that becomes your price in the answer.
Do contractor forums really affect what an assistant says about us?
They do, on the questions about whether a product is actually sound. Forums are weighed heavily because they are the one source with nothing to sell, and site experience is exactly what a specifier is looking for. You cannot edit them, but you can make the checkable version of your claim easy to find, which is what a forum reply reaches for.

See where you stand in AI answers

Run your real questions across ChatGPT, Gemini and Google's AI answers, free for 14 days. No connections required to start; all five AI engines on every plan.