Keyword Research for a Small App in Twenty Minutes

The exact method I use, including the step everyone skips — and the 90,500-a-month keyword I decided not to target, plus the 9,900 one I threw away entirely.

I do keyword research for my own apps roughly every few months, and I've landed on a method that takes about twenty minutes and consistently beats the hour I used to spend clicking around in tools.

It's not sophisticated. Most of it is knowing which step to skip and which step nobody does.

Skip the keyword discovery tools

This is the part I got wrong for a long time. Every SEO tool has a "discover keywords" feature: give it a seed, get hundreds of suggestions.

For small apps this produces garbage, and in a specific, predictable way. I gave one a seed containing the token "app" — because my product is an app — and it happily expanded on that token. Back came onlyfans app, grindr app, ad blocker. Hundreds of rows, all technically related, none describing a person who wants my product.

These tools work by finding words that co-occur with your seed. If your seed contains a generic word, that word dominates the expansion and you get a map of the internet rather than a map of your customers.

So I stopped using them and I hand-write the candidate list instead. It takes ten minutes and it's better.

Write down how people describe the problem, not the product

Sit down and write thirty phrases. The rule: describe the pain, not the solution. People search for their problem long before they know a category of software exists.

For an app blocker, the product-shaped phrases are the obvious ones — app blocker, screen time app. But the problem-shaped ones are where the volume actually is: dopamine detox, how to stop doomscrolling, digital detox, screen time passcode.

A few prompts that generate good candidates:

  • What would you have typed before you knew tools like this existed?
  • What's the specific failure? (screen time passcode — people search this because their own limit stopped working.)
  • What's the named trend? (dopamine detox is a whole movement whose participants are exactly my users.)
  • What's the how-to version? (how to block youtube on iphone.)
  • What do people call it wrong?

Include the terrible ideas. They cost nothing to check and the checking is what tells you they're terrible.

Price the whole list in one call

Then I take all thirty and get Google Ads volume, competition and CPC for the lot in a single batch. I have this wired to an MCP server so I can just ask for it in a conversation, but any keyword planner does the same job.

One batch, not thirty lookups. The point of writing the list first is that pricing it is instant.

What comes back is usually surprising. From today's run:

  • dopamine detox — 18,100/mo, competition index 0
  • how to stop doomscrolling — 1,900
  • how to stop scrolling on my phone — 10

Those last two mean the same thing to a human. One has 190× the searches. There is no way to reason your way to that from the armchair, which is the entire reason you check rather than guess.

Read CPC as a value signal, not an ad cost

This is the trick I'd most want to hand to someone starting out. I don't run ads. I still read CPC on every keyword, because CPC is the market telling you what a visitor is worth.

It's the aggregate of everyone who does run ads and does measure conversions. If they're paying $29 a click, they know something about that traffic.

Today: personal crm app has 110 searches a month at $29.51 CPC. adult friendship has 90,500 searches at $2.94.

The 110-a-month term is a better business target than the 90,500-a-month one, and volume alone would tell you the exact opposite. One is people looking to buy; the other is people reading essays. So I write for both — but I know which one is which, and I don't judge them by the same metric.

Check the SERP before you commit — the step nobody does

Here's the one that saves the most wasted work. Before writing anything for a high-volume keyword, look at what actually ranks for it right now.

Example. My relationship app could obviously target friendship app — 9,900 searches a month, right in the category name. Free money.

I looked at the results. Every single one is a make new friends app: Hoop, FriendMatch, "best apps to make friends." My app is for maintaining relationships you already have. If I ranked there — which would take months — I'd get thousands of visitors looking for something I deliberately don't do. They'd bounce, and I'd have spent the effort teaching Google that my page belongs to a category it doesn't.

So I threw away the highest-volume keyword in my category, on purpose.

Same check on adult friendship, 90,500/mo: the results are essays, Reddit threads, a psychology paper. That's not a product query — it's a reading query. So it becomes a top-of-funnel piece written like an essay, and I don't expect it to convert. Knowing that in advance is the difference between a post that underperforms and a post doing exactly what it was designed to do.

Volume tells you how many people. The SERP tells you which people. Skip the second check and you'll do months of work for the wrong audience.

Sort into three piles and write one post each

After pricing and SERP checks, everything falls into three groups:

Money terms — low volume, high CPC, buying intent. Small traffic, high value per visit. Comparison and "how to choose" posts.

Pillar terms — high volume, informational. Essays. Judge them on reach, not conversions.

How-to terms — the sweet spot. Someone with a specific problem, mid-volume, low competition, and your product is a genuine answer. how to block youtube on iphone at 1,300 a month is worth more to me than a lot of things ten times its size.

Then: one keyword per post. Two posts targeting the same phrase compete with each other, and the one that wins is decided by Google rather than you.

Then ignore all of it and write for the person

The research decides what to write about. It should have no influence at all on how you write it.

The fastest way to waste good keyword research is to produce the thing that currently ranks: a listicle with the phrase repeated eleven times, a paragraph of throat-clearing before the answer, and no actual position on anything. That content ranks for a while and gets replaced, because nobody finishes it and nobody links to it.

My rule is that once I've picked the keyword, I put the spreadsheet away and write the piece I'd want to find. Say the true thing, including the parts that don't flatter your own product — in my case, that Apple's Screen Time can be bypassed in ways that make it near-useless for adults, which is both genuinely helpful to know and the reason my app exists.

Useful and optimised aren't opposites. But when they conflict, being useful is the one that compounds.