How a SaaS company gets recommended by AI assistants

For software, the two things that most reliably get you named are public pricing and a clear, checkable description of what the product does and does not do. Assistants recommend what they can describe confidently, and a product with hidden pricing and vague scope is very hard to describe confidently.

What buyers of software actually ask an assistant

Almost never your brand name. The questions come in the following six shapes, and each one is a different page.

• Options with constraints: the best tool for a team of a certain size, in a certain country, at a certain budget, in a certain industry.
• Alternatives to a named product they already use and are unhappy with.
• Price reality checks: what this kind of software should cost, whether a quote is reasonable, what the hidden costs are.
• Can it do this one specific thing: one integration, one compliance requirement, one workflow.
• Buy or build, which people genuinely ask an assistant and almost no vendor answers honestly.
• What goes wrong: migration pain, limits people hit later, what buyers regret.

Read that list again and notice how few of them your website answers. Most software sites are organised by feature, and none of these questions is about a feature.

Which sources the assistants read for this trade

For software questions the mix is fairly consistent. The following five do most of the work, in roughly this order of influence.

• Roundup and comparison articles, which supply the candidate list for any "best tool for" question.
• Software directories and review platforms, which carry your category, your price band and a description you may never have read.
• Your pricing page, if it has a price on it. This is the single most quoted vendor page in the category.
• Your public documentation, which answers the "can it do this one thing" questions better than any sales page and is quoted with more confidence.
• Community threads, which own the "what goes wrong" questions completely.

Two of those five are pages you own, and both are pages most software companies neglect in favour of the ones nobody cites.

Why hidden pricing costs you so much

When somebody asks an assistant for options in a category, the answer is more useful if it can say what things cost. A product with a public price can be placed in that answer. A product that says "contact us" often gets a sentence noting that pricing is not published, or gets left out in favour of one that can be named.

We have a dated instance of what not publishing actually costs. On 6 August 2026 the pricing page for clawlaw.in, our sister company, rendered its prices only after scripts had run, so what a crawler received contained no prices at all, and the prices the assistants did quote had come from an app store listing instead of from the company's own site. On the same day ChatGPT noticed that the website and the app store listing carried different plan names and different prices for the same product, and said so in its answer. An unpublished price does not stay unpublished. Somebody else supplies it, and their version may not agree with you. If you cannot publish every price, publish a starting price and what changes it.

The internal objection is usually that pricing is complicated and a number will be misread. The answer is to publish the number and the conditions together, which is what a buyer wanted anyway, and to date it.

The three fixes that matter most here

One: put a number on the pricing page, with a date. A starting price, what it includes, the three or four things that change it, and when you last reviewed it. This single page will be quoted more than anything else you own.

Two: make your documentation public and indexable. Docs are specific, they rarely exaggerate and they answer the narrow questions buyers actually ask. If yours sit behind a login, your most citable asset is invisible. This is usually a configuration decision rather than a writing project, which makes it the cheapest win available.

Three: get into the directories and roundups that rank for your category. For broad questions the candidate names come from those pages, and a complete, current listing there affects several questions at once. Your absence is a ceiling that no amount of writing lifts.

After those three, a fourth is worth naming even though it is harder: a real scope statement. Which integrations, which limits, which platforms, named. "Works with your existing stack" is unusable. A list is quotable.

A worked example

On 17 September 2026, on a question about finding every case filed against a company, Claude ranked two enterprise vendors above clawlaw.in, our sister company. The reason it gave was that those two vendors publish explicit court and tribunal coverage lists. It added that its ordering reflected price transparency and source authority rather than product quality.

Three things in that are worth taking, and all three apply to software.

The first is that the winning asset was a list. A published statement of exactly what is covered and what is not, which a reader can check line by line. In software that is your integration list, your supported platforms, your regions, your export formats and your limits, written as text rather than drawn as a grid of logos.

The second is that the assistant named price transparency in the same breath. A product whose price can be stated can be placed in an answer. A product that says contact sales often gets a sentence noting that its pricing is not published, or gets left out in favour of one that can be named.

The third is the uncomfortable one. The answer said the ordering was not about product quality. So the gap between being named and not being named was not the product. It was what each company had published about its own scope and its own prices, and closing that gap is a week of work rather than a development plan.

Turn it into a check on your own site. Take one narrow buyer question, the sort with a constraint in it: a team size, a country, and one specific channel or integration. Then ask three things. Is your name on any page that lists tools for that constraint, because that is where candidate names come from. Does any page of yours state, in text, that you support the thing in the constraint, and since when. And can an answer say what you cost. If your three answers are a feature grid, a logo and a demo form, an assistant cannot produce a complete sentence about you, and a complete sentence is what gets included.

What does not work

Thought leadership on trends nobody is asking about. Comparison pages that conclude you win every category, which are read as advocacy and discounted. Claims with no source, especially about competitors' prices, which is one of the fastest ways to have your whole page dismissed. Feature pages written for a keyword rather than a question. And case studies with no numbers in them, which read as decoration.

What this approach cannot do

It cannot make you the answer to a question you do not fit. If you sell to enterprises and the question is about free tools for freelancers, the honest answer names somebody else, and that is the engine working properly.

It cannot beat the analyst and large publisher sources on the broadest questions in crowded categories. Narrowing is the route, not persistence.

It cannot deliver results in a fortnight. Directory listings take time to be crawled, and roundup inclusion depends on other people's schedules.

And it cannot be guaranteed. No one can promise you a place in an AI answer, and anyone who does is selling you something they do not control.

How to measure it

Take the six question shapes above, write ten to fifteen real questions in your buyers' words, and ask them on two or three assistants without naming your company. Record whether you were named, who else was, and every source cited. Repeat monthly with the same set.

Watch two numbers separately: how often you are named, and how often any page of yours is cited. The second one moving while the first does not is the most common pattern, and it tells you the mechanics work and the pages are not quotable yet.

Common questions

We genuinely cannot publish pricing. What then?

Publish the shape of it. A starting point, the units you charge by, the things that increase it, and a worked example for a company of a given size. That is enough for an assistant to describe you, and it is more than most of your competitors offer.

How long before this shows up?

The retrieval based engines can reflect new pages within days of them being findable. Engines that answer partly from memory move on a much longer and unpublished timescale. Measure on the fast ones and be patient with the rest.

Do we need reviews on review sites?

Where those sites are cited in your category, yes, and the source tally from your own measurement tells you whether they are. Do not assume. In some categories directories matter far more than review platforms.

An assistant cited a competitor's comparison page for a fact about our product. What do we do?

Check the fact, request a correction if it is wrong, and keep a dated copy of the page. Then fix the underlying cause, which is that the fact was not clearly stated on a page of yours. Facts inside screenshots, feature grids and PDFs are the usual reason a competitor ends up as the authority on your own product.

Is being cited the same as being recommended?

No, and in software the gap is common because vendor pages are used for prices and features while the recommendation comes from a roundup. Track both. Citations tell you the mechanics work, recommendations tell you the evidence and the fit persuaded.

Should we build a page for every competitor comparison?

Only where you can be honest, including about what the other product does better. A comparison page that wins every row is discounted on sight. A page that concedes two rows and explains who should buy the other product is the rare one that gets quoted.

What to do first

This week, do the two cheapest of the three fixes. Put a number and a date on your pricing page, and take your documentation out from behind the login if it is there. Both are decisions rather than projects, and together they change what an assistant is able to say about you at all.