Building Discoverability in Unknown Topics: How Software Companies Are Building Recognition for What They Do

Let’s start with a pretty obvious fact: people can’t search for something they don’t know exists.

That sounds like a small problem until you’re the marketing team at a software company that just built something new. Your product solves a real problem. You gave the category a name. The name is good. And the search volume for that name is zero.

TL;DR

• New software categories start with zero search volume, and both Google and LLMs are built to answer the questions people already ask. Nobody asks about a category they don’t know exists.
• We tracked three B2B SaaS companies launching new categories with similar budgets. The one that focused on defining its category won 62% of “what is” prompts, but showed up in only 5% of the problem questions buyers ask.
• Problem-first content got the category term searched after 5 to 6 months. The market learned the name from the answers.
• Problem content plus third-party mentions (G2, podcasts, partner docs, cited research) reached 46% AI visibility and 51 demos a month. The category-first approach got 3.
• The metric to track is how often AI names your category without being asked. I saw the same pattern launching the first prepaid card in Chile at MACH.

Most teams I talk to respond the same way. They publish the definitive guide. The manifesto. The “What is [our category]” page. They try to own the definition before anyone else does.

I know this situation well because I lived it. I joined MACH, a Chilean fintech, as one of its first hires, leading content and brand. After a lot of research, we launched the first prepaid card in Chile.

The problem was that nobody searched for “prepaid card”. It was a new thing, so nobody knew to look for it, and I quickly realized I had no obvious way to sell the product or get traction. What people did search for was their problem: how do I pay for Uber or Spotify if I don’t have a credit card?

So that’s what we wrote about. That problem-first content turned organic into MACH’s main acquisition channel, and helped put a MACH account and card in the hands of 1 in 10 people in Chile. If you only count the people who could actually sign up (kids can’t open an account), it’s closer to 1 in 4. For a product whose category had zero searches the day we launched, that still surprises me when I say it out loud.

So when I see a new category launch with a manifesto and a definitive guide, I get a little nervous. But one story is one story, and AI search changes how discovery works.

We wanted to see whether the same pattern holds today, so we looked at three B2B SaaS companies that launched new categories. Similar starting points, similar budgets, three different content strategies, tracked for 12 months in Google and in LLMs like ChatGPT and Claude.

The company that won the definition of its own category ended up with 3 demos a month. The one that focused on the buyer’s problem and got other sources to back it up ended up with 51.

Why is discoverability so hard when your topic is new?

Search engines and LLMs are built to answer questions people already ask. That works well for established categories. If you sell CRM software, there are thousands of people typing “best CRM for small business” every month, and your job is to show up for them.

New categories don’t have that. They have three problems stacked on top of each other:

• No search volume. Google has nothing to rank you for, because nobody types the term.
• No shared vocabulary. Buyers describe the pain in their own words (“reconciling NetSuite and SAP takes my team a week”), and none of those words is your category name.
• No reference point for AI. An LLM builds its understanding of a topic from what many sources say about it. If the only source talking about your category is you, the model has very little to work with.

And there’s a fourth problem that’s easy to miss. LLMs respond to prompts. They don’t go looking for good content to promote. If nobody asks about your category, being the best source on it doesn’t create exposure, because the question never gets asked.

So the real question for these companies is how to get a market to start asking a question that doesn’t exist yet.

Who did we track, and how?

We looked at three B2B SaaS companies, each launching a category their market didn’t have a name for yet:

The starting points were very close:

• Search volume for the category term: 0/month
• Branded searches were similar across all 3 competitors. 
• AI mentions: 0%
• Content plan: roughly 24 pieces over 12 months (with a 15% difference across the board)

What changed was how each one spent those 24 pieces.

• Company 1 published pieces about its category: what it is, why it matters, a manifesto, the definitive guide.
• Company 2 published pieces answering what its buyers already search for, presenting the category as the answer.
• Company 3 published mostly problem-focused pieces and used the rest of its effort on external appearances: G2, comparison lists, podcasts, partner integration docs and a report cited by industry media.

We tracked visibility on two fronts every month. In Google, we followed search volume for each category term and branded searches. In LLMs like ChatGPT and Claude, we ran a fixed set of 50 prompts per company, built around their own market:

• 30 problem prompts, in the words buyers use. For Company 1: “how do I reconcile transactions across NetSuite and SAP”. For Company 2: “why do we have overstock in one warehouse and stockouts in another”. For Company 3: “how do I cover last-minute shift gaps across 40 stores”.
• 15 category prompts, like “what is agentic reconciliation”.
• 5 comparison prompts, like “best tools to agentic tools to process month-end close”.

Let’s get to the data. What happened after 12 months?

AI mentions in problem prompts (the ones buyers ask)

Month	Company 1 (category first)	Company 2 (problem first)	Company 3 (problem + third parties)
0	0%	0%	0%
3	1%	6%	9%
6	2%	14%	22%
9	4%	22%	35%
12	5%	28%	46%

By month 12, Company 3 shows up in almost half of the answers to real buyer questions. Company 1 shows up in 1 out of 20.

AI mentions in category prompts (“what is [category]”)

Month 12	Company 1	Company 2	Company 3
Named in the answer	62%	31%	48%

This is where Company 1 wins, and by a lot. When someone asks what agentic reconciliation is, Company 1 gets named 62% of the time. The catch is that very few people ask that question. Company 1 owns 62% of a prompt with almost no demand.

Search volume for each category term (market-level searches per month)

Month	Company 1	Company 2	Company 3
3	10	0	20
6	20	40	90
9	30	110	220
12	40	210	390

Company 2’s category term didn’t register at all until month 5 or 6. Then it accelerated. Company 1, the company that published only about its category, ended the year with the lowest search volume for it (40 searches a month).

Category “installation”: how often AI uses the category name without being asked

This one needs a quick explanation. We looked at the answers to problem prompts and counted how often the AI used the company’s category name when the user never mentioned it. In other words, how often the model offered the category as the answer to a problem.

Month 12	Company 1	Company 2	Company 3
Spontaneous use of the term	1%	12%	24%

This is the moment a category starts to exist for AI. Up to that point, the model can define the category when someone asks. From then on, it recommends the category to people who didn’t know it existed.

Business impact at month 12

Metric	Company 1	Company 2	Company 3
Branded searches	+12%	+52%	+105%
AI-referred sessions/month	40	420	780
Demos attributable to organic + AI/month	3	28	51

Similar budgets, similar starting points. Company 3 doubled its branded searches and got 17 times more demos than Company 1.

Why did each strategy perform this way?

Company 1: great answers to a question nobody asks

Company 1 did what most category creators do, and it did it well. Its content became the best reference on their category out there, which is why it gets named in 62% of category prompts.

The problem is where the demand sits. LLMs answer the questions people ask, and finance teams ask about their problem (“why does month-end close take us two weeks?”). Company 1’s content doesn’t answer that question, so it doesn’t show up. Google works the same way. 

Company 2: entering through the questions that already exist

Company 2 went to where its buyers already were. Its content answers “why do we have overstock in one warehouse and stockouts in another” and then says, more or less, “the way to solve this is predictive stock rebalancing”.

Repeat that across different pieces and the AI starts learning an association: problem → category → brand. After a while, the model starts repeating it on its own. That’s why the category term started getting searched around month 5 or 6. People read the answer, learned the name, and then went to search for it.

Company 3: fewer pieces, more corroboration

Company 3 published 8 fewer pieces on its own site than the other two. It still won by a wide margin.

LLMs trust claims more when several independent sources say the same thing. When G2, a comparison list, a partner’s integration docs and an industry outlet citing Company 3’s report all connect “covering shift gaps” with “frontline shift intelligence” and with Company 3, the model has much more reason to believe it than when a company says it 24 times on its own site.

The result: roughly 1.6x Company 2’s AI mentions (46% vs. 28%), and almost double the demos (51 vs. 28), with a third less owned content.

What are the five findings that matter?

1. Owning a definition that nobody searches for doesn’t give you visibility. Company 1 wins 62% of category prompts and still ends up 9x below Company 3 in the AI answers buyers see (5% vs. 46%).
2. Category demand comes after the content. The category terms only started growing after problem-focused content introduced them. Company 1, which talked about its category the most, ended with the least search volume for it.
3. There’s a lag of about 5 to 6 months. Before that, problem-focused content looks like it isn’t working if you only track the category term. This is the point where most teams give up and switch strategies (usually to writing the manifesto).
4. Third-party validation beats volume. 16 owned pieces plus 8 external mentions outperformed 24 owned pieces on every metric we tracked.
5. Look for spontaneous use of the term. Track how often AI brings up your category in answers to problem questions, without the user mentioning it. When that number starts moving, the category exists.

So how do you build discoverability for a category nobody searches yet?

If you’re launching something new, this is how I’d split the work based on what we saw:

1. Map the problem language first. Before writing anything about your category, list the questions your buyers already ask. Use sales calls, support tickets, G2 reviews of adjacent tools and search data for the pain point. Those questions are your real keyword list.
2. Answer the problem and name the category inside the answer. Every problem-focused piece should explain the pain, explain why current approaches fall short, and introduce your category name as the solution. Consistently, with the same term every time.
3. Spend some of your budget outside your site. Review platforms, comparison lists, podcasts, partner documentation, original research that media can cite. You need other sources saying what you say, because that’s how LLMs gain confidence.
4. Keep one solid definition page. You still want to own “what is [category]” for when people start asking. One strong page is enough; you don’t need 24.
5. Measure spontaneous use, not just category volume. Run a fixed set of problem prompts every month and track how often AI names your category without being asked. It moves before search volume does.
6. Plan for the 5 to 6-month lag. Tell leadership upfront that the category term won’t register for months. If the expectation is set, the team doesn’t abandon the strategy right before it starts working.

Where does a new category really start?

It’s tempting to think a category starts the day you name it and publish the definition. In our data, the categories started existing somewhere around month 5 or 6, when an LLM answered a buyer’s problem question and brought up the category on its own.

That happened because the companies that won spent their time where buyers already were. They answered the questions people were asking, used the same category name every time, and got other sources to repeat it. The market learned the term from those answers, and only then started searching for it.

If you’re about to launch a category, a good first step costs nothing: pull your last 20 sales calls and write down how prospects described the problem before they knew you existed. That’s where your content should start.