Most AI search advice for software companies ends at the same place: buyers start with a problem, not a product, so write content that answers the problem. Show up when someone asks how to fix the thing your product fixes, and you will be in the conversation before the shortlist exists.
The first half is true. The second half assumes that being the source of the answer gets your product named in it. I wanted to see whether it does.
In November I asked Google AI Mode three problem questions, written the way someone at a software company would ask them, with no product named in any of them. Our booked sales demos have a 30 percent no-show rate, how do we fix it. Our support team is drowning in repetitive tickets, what should we do. Our month-end close takes 15 days, how do we get it under a week. Then I read every source behind each answer and checked which companies were named as something to use.
Software vendors wrote most of the answers
In all three, most of the cited pages were blog posts by software companies selling into the problem. The demo answer drew on posts from a sales meeting platform, a scheduling and routing tool, a gifting platform and several others. The support answer cited posts from help desk vendors, AI support agent startups and a customer portal provider. The month-end answer cited posts from a spend platform, a close management tool, two billing and revenue software vendors, a planning software vendor and a finance automation vendor, alongside a few consultancies.
So the advice works, up to a point. Write a useful post about the problem and the engine will read it, quote it and link it. Most of the advice in these answers came from vendors.
Most of them were not named
The answers did name products, but not in proportion to who wrote them.
The demo answer told the reader to cap how far ahead prospects can book, using a scheduling tool “such as Calendly or Chili Piper”. Neither company had a page cited. It suggested a short product video using “tools like Loom or Walnut”. Walnut’s LinkedIn post was cited. Loom’s was not. The sales meeting platform and the scheduling tool whose posts were cited for the main advice were not named anywhere in the answer.
The support answer went the other way. Three of the vendors whose posts it cited were named as the thing to use: Bitrix24 for a customer portal, and Pylon and Mosaic AI as AI agents to train on past tickets. The help desk vendors whose posts it cited for tagging and knowledge base advice were not named. The big help desk brands appeared only in a follow-up question asking which tool the reader already used.
The month-end answer named four products. Ramp and Numeric each had a post cited and were named, Ramp for corporate cards and expenses and Numeric for close management. Brex and FloQast were named next to them with only their homepages linked. At least four other software vendors had close posts cited, and none of them was named.
What separated the named from the cited
My first guess was self-promotion: posts that talked about their own product more would get it named. I counted how often each cited post mentions its own brand. One of the named vendors mentions itself far more often than any other post I checked. The others that were named sit in the same range as the vendors that were not. Brand mentions did not explain it.
What did fit was the shape of the answer. Each answer was a plan with steps, and only some steps were “use a tool for this”. When the answer reached a step like that, it named products for it. If a cited vendor sold exactly that tool, it had a chance. If the category had famous names, those got named too, cited or not. If the vendor’s post was cited for a step that was not a tool step, like tagging tickets, changing SDR pay or setting a materiality threshold, the vendor got the credit for the advice and no mention for the product.
That matches the difference between being cited and being recommended. In these answers, writing the post earned the citation. Whether it earned the recommendation depended on whether the plan had a step your product fills.
What to do
Write problem content that ends in your product’s step. A post about cutting demo no-shows that is mostly about SDR pay and calendar invites will be cited for SDR pay and calendar invites. If your product solves one step of the problem, make that step central, explain it as a method, and name your product as one way to do it. The engine needs a tool step before it can name a tool.
Name the category of tool, not just your brand. The answers named products under labels like “close management tools” and “AI agents”. If your post never says what kind of tool you are in plain words, the engine has nothing to file you under when it reaches that step.
Expect to share the step with the famous names. Where a category has a well-known leader, the engine named it even when none of its advice was cited. Your post is how you get named next to them. It will not usually replace them. That is still worth having, because it puts you on the reader’s list at the moment they decide to buy a tool.
Map which problem questions have a tool step. Some problems in your space are mostly process. Content there will earn citations and goodwill, not recommendations. When you build your buyer query map, mark which problem questions end in “buy a tool” and put your effort there first.
Run the question and read the plan, not the sources. Most teams check whether they are cited. Check which step you are cited for, and whether that step names a product.
What I got wrong
I expected the vendors that wrote the most self-promotional posts to be the ones the engine named, and I was ready to tell you to tone it down or turn it up. The brand counts did not support either. The thing that mattered was not how loudly a post mentioned its product but whether the answer had a place to put one.
These were single runs of three questions on one engine, and Google offered to rerun them without personalisation. The brand counts were rough, taken from the page text including some navigation. Treat this as a read of the pattern, not a measurement. Run your own problem questions before you act on mine.
Where this leaves you
Problem content works: in these answers, software vendors supplied most of the advice. But being the source of the advice and being named as the tool are separate outcomes, and the second depends on the answer having a step your product fills. Write for that step, say plainly what kind of tool you are, and check which part of the plan you are being cited for.
If you want to see which problem questions in your market end in a product recommendation, and whose, that is part of the engagement, and it runs on a documented method.