Most competitor analysis starts with a list of names the founder already had. The research then confirms what the list implied. It is a thorough process for answering a question nobody needed answered.
The competitors that damage a business are rarely on that list. They are the alternatives the customer chose instead — including alternatives that aren't companies at all.
Four categories, and only one is obvious
Direct competitors sell roughly what you sell to roughly who you sell to. These are the names you already have. They matter least for research purposes precisely because you already watch them.
Category substitutes solve the same problem a different way. A business looking for a marketing tool might hire a freelancer, sign an agency retainer, or promote someone internally. None appears in a competitor analysis of software vendors, and all three win deals.
The spreadsheet. For a large share of small businesses, the incumbent is a spreadsheet plus a habit. It costs nothing, nobody has to be convinced, and it is the most common reason a deal stalls.
Doing nothing. The problem is real but not urgent enough. This is often the single largest bucket, and it never appears on a competitive matrix because it has no logo.
A competitor analysis that only covers the first category tells you how to win a comparison you're rarely in.
Where the real list comes from
Search results for the problem, not the product
Search the phrase a customer would type before they know what category of solution exists. Not "marketing automation software" — "can't keep up with posting on social media."
Record everything that ranks: tools, agencies, courses, a blog post that satisfies the need for free, a Reddit thread that talks someone out of buying anything.
That page-one set is closer to your real competitive field than any vendor list. It's what the buyer sees at the moment they're deciding.
What AI engines recommend
This one is newer and moves faster than most competitive research accounts for.
Ask ChatGPT, Claude, Perplexity and Google's AI Overviews the questions your buyer would ask: "What's the best way to handle marketing for a five-person business?" Record what gets named.
Two things come out of it. You learn whether you're in the consideration set at all — increasingly the consideration set, since a meaningful share of buyers now start here rather than with ten blue links. And you learn which sources the engines cite, which is usually a comparison article or a roundup rather than any vendor's own site.
That second finding is the actionable one. If Perplexity consistently cites a particular roundup and you aren't in it, getting into it is a concrete, achievable task with a clear outcome — and it is invisible to any research that only looked at competitors' websites.
Review sites, read for the switch
G2, Capterra and their equivalents publish comparison data most people skim. The useful part is not the star ratings; it's what reviewers say they moved from and what they nearly chose instead.
Worth knowing that G2 acquired Capterra, GetApp and Software Advice in February 2026 but kept the review pools separate. Presence on one does not mean presence on the others, and a competitor strong on one site may be absent from another — so checking a single site gives a partial picture in both directions.
Your own lost deals
The highest-quality source, and the one almost nobody systematises.
Ask every prospect who doesn't buy what they did instead. Not "why didn't you choose us" — people are polite and will invent a reason. "What did you end up doing?" Answers cluster fast, and the cluster is usually not what the founder expected.
Twenty answers is enough to reorder a competitive picture.
What to record
Once the list is real, most competitive matrices still collect the wrong things. Feature grids are easy to build and rarely change a decision.
The fields that actually inform strategy:
| Field | Why it matters |
|---|---|
| Who they say it's for | Positioning gaps show up here first |
| Entry price | Where the market's price floor sits |
| What's deliberately excluded | Reveals their strategy more than the feature list |
| Words used for the problem | The vocabulary your buyers already have |
| Proof offered | Case studies, numbers, named customers |
| Where they're cited | The pages that feed AI recommendations |
"What's deliberately excluded" repays the most attention. Everyone lists what they do. Very few competitors do everything, and the consistent gap across a whole field is a market position sitting unclaimed.
Where AI helps and where it doesn't
Language models are strong at reading a large corpus and finding structure in it. They are unreliable at recalling facts about specific companies.
Works: paste twelve competitors' pricing and homepage copy in, ask for a comparison, ask what nobody is saying. Every conclusion traces to text you supplied.
Doesn't work: asking a model who your competitors are, what they charge, how many customers they have, or how much they've raised. All of that is produced from memory, and memory here means plausible reconstruction. Funding figures and headcounts are fabricated with particular confidence.
The rule is the same one that governs all AI research: supply the source or verify the output. There is no third option.
Reading the result
Three questions turn a competitor document into a decision.
Who is nobody serving? Usually visible in the "who it's for" column. If eleven of twelve competitors describe mid-market teams, the gap is at both ends.
What is everybody claiming? A claim every competitor makes has no persuasive value. It's table stakes, and repeating it costs you your differentiation while sounding safe.
What is nobody willing to admit? Every category has a limitation all vendors soften. Stating it plainly is uncomfortable and disproportionately effective, because it is the one sentence a buyer cannot get anywhere else — and it is the kind of specific, first-hand judgement that generic content can never contain.
How often
Quarterly is sufficient for most markets, with one exception: the AI-engine check is worth running monthly. Those recommendations shift as sources get published and re-crawled, and a change there can happen without any competitor doing anything visible.
Questions
How many competitors should I analyse?
Eight to twelve, spread across all four categories. A list of twenty direct competitors is less useful than a list of eight covering direct, substitute, spreadsheet and nothing.
Can AI tell me who my competitors are?
Not from memory — that output is unreliable. But asking AI engines what they'd recommend for your customer's problem is a legitimate research method, because you're recording what a real buyer would be shown rather than asking for a fact.
What about competitor traffic and revenue estimates?
Third-party traffic tools are directional at best and often wrong by large multiples for smaller sites. Useful for spotting trends, not for planning against.
How do I compete with "doing nothing"?
Different work entirely: it's about establishing cost of inaction rather than superiority. Comparison content doesn't reach these buyers; content about the problem does.
Should I mention competitors by name in my content?
Comparison pages tend to rank and are frequently cited by AI engines, which favour them for recommendation queries. The constraint is accuracy — a claim about a competitor needs to be current and defensible, and pricing pages change without notice.