How to Do Keyword Research for B2B SaaS in 2026

How to do keyword research for B2B SaaS in 2026

Keyword research did not die, it moved. The job used to be: find a high-volume head term, write a page, collect the traffic. For B2B SaaS in 2026 that playbook is mostly wasted effort, because the broad informational terms it chased are exactly the ones AI Overviews now answer on the results page without sending a click. So modern keyword research starts at the bottom of the funnel, where buyers still click and still convert, and works back up from there.

This is the actual process I run on B2B SaaS accounts, step by step. It is intent-first, money-keyword-first, and built for a world where ranking the page is only half the job and being named in the AI answer is the other half.

Volume tells you how many people search. Intent tells you whether any of them are your buyer. Only one of those pays.

How AI changed keyword research

Old keyword research worked the top of the funnel. The goal was reach: rank a big “what is” or “how to” term, pull in eyeballs, and trust that some of them trickle down to a sale. That made sense when the only way to get an answer was to click a result.

It does not make sense now. AI Overviews capture the click on broad informational queries, so chasing those big head terms increasingly means winning impressions you never convert, the exact split behind the Great Decoupling. No one is clicking the top of the funnel anyway, so pouring marketing dollars into ranking there is spending into a headwind.

The reallocation is the whole point: take the energy you used to spend chasing volume and put it on the bottom-of-funnel terms where the click still lands and the buyer is ready to spend. Volume becomes a ceiling you note, not a target you chase. Intent, winnability, and whether a query still sends a click become the real filters. The four steps below are how you find those terms and turn them into a plan.

Step 1: Identify your money keywords

Money keywords are the bottom-of-funnel, commercial-intent terms your buyers search in the buying moment. For B2B SaaS they cluster into a few recognizable shapes:

• Category terms ending in software, platform, provider, or solution: “legal case management software,” “XDR platform,” “employee recognition solution.” A searcher typing these is shopping, not learning.
• Comparison and “vs” queries: “[your product] vs [competitor],” “[competitor A] vs [competitor B].” These are buyers down to a shortlist.
• “Alternatives” queries that name you or a competitor: “[competitor] alternatives,” “[your product] alternatives.” These are buyers actively switching.
• Pricing, demo, and ROI terms: the searches only a near-purchase buyer runs.

Go wide first. Map every money keyword relevant to your specific niche, not just the obvious category term. These are the queries that decide deals, and they are where the rest of the process points.

Step 2: Cut it to a 50-keyword shortlist

You cannot chase everything, and you should not try. Narrow the list to roughly 50 money keywords, max, using one blunt question on each term:

If we ranked number one, or got named in the AI Overview and the LLMs, for this term every single time, would we have drastically more revenue?

If the honest answer is “hell yeah,” it belongs on the shortlist. If it is “maybe, a little,” it does not. That filter keeps you focused on the terms that actually move the business instead of a 600-row spreadsheet you will never action.

One corroborating signal worth checking: money keywords usually carry a high cost-per-click in paid search. That CPC is the market telling you advertisers have already proven the term converts, which is exactly why it is worth your organic effort too. Then weigh that value against keyword difficulty (KD): a term with low KD and high CPC is the best target there is, cheap to win and worth winning.

Step 3: Read the SERP and audit your content

Now take the shortlist and, for each term, run two checks.

First, do you already have a page targeting it? You will often find you are half-covered: a page exists but is thin, aimed at the wrong intent, or buried.

Second, what is actually ranking, and in what format? This is the step most people skip and the one that decides everything. Pull up the live results page and read the winning page type. If your competitor is winning a term with a platform page and you are pointing a blog post at it, you do not need a better blog post, you need a better platform page. If a listicle owns the SERP and you are aiming a product page at it, you need a listicle. Match the format Google is already rewarding instead of insisting yours is better, which is the same do-not-fight-Google discipline that intent research runs on.

While you are in the SERP, note whether an AI Overview fires and how much organic click is even left to win. And look for striking-distance terms: money keywords your site already ranks for at roughly positions 5 to 20, with impressions but few clicks. Those are the fastest wins on the board, because you are one improvement away from page one rather than starting cold.

Step 4: Map the gaps and build the plan

With the SERP read and the content audit done, every money keyword sorts into one of two actions:

1. Refresh (retarget) an existing page. You already have a page in the neighborhood, so you improve it and re-point it at the right intent and the right format. Refreshing and retargeting are the same move: taking a page you already own and aiming it correctly. A blog post fighting a SERP that rewards platform pages is a retarget too.
2. Build net-new. You have no page for the term, or no page in the format the SERP rewards, so you create one.

That triage is your content plan. Sequence it by revenue impact, not by what is easiest, the same way you would prioritize anything by influenced pipeline. The output is not a keyword list, it is a ranked queue of pages to refresh or build, each tied to a term you already decided was worth real money.

The four steps on a real SERP: a project management SaaS

Theory is cheap, so here is the method run end to end on live data. Say you run SEO for a mid-market project management SaaS, up against Monday, Asana, and ClickUp. Here is what Steps 1 and 2 surface when you pull the money keywords (real US volume and cost-per-click):

Money keyword	Volume	KD	CPC
project management software	135,000	51	$44.27
project management tool	22,200	56	$38.84
asana vs monday	1,600	5	$17.52
monday.com alternatives	480	0	$41.12
clickup alternatives	480	0	$32.18
asana alternatives	480	1	$28.58
best project management software for small teams	390	28	$53.07
project management software pricing	170	21	$24.61

Read the three data columns together, because each answers a different question. Volume sizes the prize. Cost-per-click confirms commercial value: “best project management software for small teams” gets only 390 searches a month, but advertisers pay $53 a click, because the people typing it are buying. And keyword difficulty (KD, scored 0 to 100) tells you how hard the term actually is to rank for, which volume alone never does.

Put the columns side by side and the play is obvious. The head terms are KD 51 and 56, genuinely hard, and an AI Overview is skimming their clicks anyway. The “alternatives” money keywords are KD 0 to 1 with $28 to $41 CPCs: almost no ranking difficulty and high commercial intent at the same time. Low difficulty plus high commercial value is exactly the combination you are hunting for, and a volume-first researcher walks right past it to chase the KD-51 head term.

Step 3, the SERP read, is where the plan actually forms. Three terms, three different stories:

“Project management software” (135,000 searches, KD 51, AI Overview fires). The top ten is almost all listicles (“Top 10 Project Management Software,” “20+ Best…,” “Honest Review of 6…”) plus Reddit, YouTube, and Wikipedia. The only product pages that crack it are ProjectManager.com, whose brand name is the keyword, and Microsoft. Read that honestly: you are not ranking your platform page number one here, and an AI Overview is skimming clicks off the top anyway. High volume, low winnability. Deprioritize it as a direct target.

“Monday.com alternatives” (480 searches, KD 0, $41 CPC). A completely different SERP, and the most actionable one. The number-one result is Teamwork.com’s own “10 best monday.com alternatives” page: a direct competitor winning a competitor’s alternatives term with a comparison listicle. Monday’s own alternatives page sits at number three, and Baserow, Airtable, and others run the same play below. The winning page type is unmistakable, a vendor comparison page that positions itself as the alternative, and it is one you can build and win.

“Best project management software for small teams” (390 searches, KD 28, $53 CPC). Reddit takes number one, and the rest is listicles and YouTube tier-lists. A product page has no place here; the format Google rewards is an experienced “best for small teams” roundup. To compete you need a genuine, methodology-backed listicle or a mention inside the AI Overview, not a feature page.

Step 4 turns those reads into a content plan, sequenced by winnable revenue:

1. Build the “[competitor] alternatives” pages first (monday.com, ClickUp, Asana). At KD 0 to 1 they are nearly unguarded, they carry $28 to $41 CPCs, they catch buyers mid-switch, and the SERP proves vendors win them. Format each as a comparison page with first-hand experience and a verdict, the kind a model will name rather than summarize, positioning you as the top alternative.
2. Build a “for small teams” segment listicle, not a product page, because that is the format the SERP rewards. Give it a real methodology so it holds up.
3. Do not chase the head term with a platform page. “Project management software” is a get-named play: earn citations in the third-party listicles that rank and structure the content to be pulled into the AI Overview, rather than burning months trying to rank a feature page the SERP does not want.
4. Pull the prompt list straight from the SERP. The People Also Ask box (“What is the most popular project management software?”, “Who is Monday.com’s biggest competitor?”) and the related searches (“monday.com alternatives reddit”) are your starting set of prompts to track, the bridge to prompt tracking below.

Same eight keywords, three different SERP shapes, one plan that targets the winnable, high-intent terms instead of the vanity head term. That is the whole method in practice.

Do it faster with Ahrefs and Claude

The four steps are the method; tooling is how you run them quickly. The workflow I use has three parts, because no single tool does the whole job:

1. Pull the data from a real tool. Ahrefs or Semrush for volume, difficulty, CPC, the SERP, and the related-term and question expansions. This is the raw material, and you need a genuine data source because the next step has none of its own.
2. Do the judgment in Claude. Hand the export to a model to cluster terms into topics, flag where your existing pages cannibalize each other, and prioritize by intent. Connecting Ahrefs to Claude over its MCP server is what made this fast for me: the data lands in the conversation and the clustering that used to take a day happens in one pass.
3. Verify the live SERP by hand. Rankings lag in every tool and AI Overviews refresh nowhere in real time, so you run the important queries in Google yourself before committing a term to the plan.

The tool gives you a list. The model turns it into priorities. The manual check keeps you honest.

Which keyword research tools should you use, free and paid?

You do not need an expensive stack to start; you need the right tool for each job. The honest landscape:

• Google Keyword Planner (free). The classic starting point, built into Google Ads. It pulls straight from Google and costs nothing, but it is made for advertisers, so it reports volume in broad ranges and skews commercial. Good for a first pass and a volume sanity-check, not precise enough to prioritize a whole plan on its own.
• Google Search Console (free). The most underrated keyword tool, and you already own it. It shows the exact queries you already rank for, including the striking-distance terms on page two that a refresh can push onto page one. Start here before you buy anything.
• Ahrefs or Semrush (paid). Worth it once you are serious. Real volume, keyword difficulty, CPC, SERP data, the related-term and question expansions Keyword Planner hides, and the competitor gap analysis free tools cannot do. This is the data source the four-step method above assumes.
• Ubersuggest and free tiers (budget). A cheaper middle ground with lighter data. Fine when Ahrefs is out of reach, as long as you treat the numbers as directional.

The tool matters less than the method. Mine the live SERP for intent first, because People Also Ask and related searches are free and current, then use whichever tool you have to attach volume, difficulty, and CPC. A paid tool makes you faster; it does not make you right.

When you have exhausted the money keywords

There is a finite supply of money keywords. Once you have targeted them all, and ranked or gotten named for the ones you can win, there is only so much more bottom-of-funnel capture to do. That is when a keyword program has to mature.

The next layer is demand you create rather than demand you capture: use-case pages, pain-point content, industry and vertical pages, and genuine thought leadership built on a point of view and first-party data. This is a maturing SEO strategy, expanding outward from the buying-moment terms into the surrounding problems your buyer has before they know your category exists. Do it in that order, though: own the money keywords first, because that is the revenue you can attribute, then branch out.

From keyword tracking to prompt tracking

The bigger shift underneath all of this: keyword tracking is becoming prompt tracking. It is no longer enough to know your rank for a term; you need to know whether your brand is named when a buyer asks an AI tool about your category. That is what tools like Scrunch and Profound exist to measure, and it is fast becoming part of the research job, not a separate one.

So alongside your keyword shortlist, build a shortlist of prompts your target market could be searching. The catch, and it is a real one: ChatGPT and AI Overviews do not give you click or search volume on long-tail prompts the way a keyword tool gives it on keywords. The volume data is still keyword-based, so you cannot size prompts precisely yet. You have to do your best.

The way to do your best is to pair your pain points with your keywords into the long-tail question variants a buyer would actually type, and to mine People Also Ask for the real phrasings people use. That gives you a defensible set of prompts to track even without volume on each one. Expect the data to improve: knowing exactly what people prompt is a privacy question for now, but the platforms will almost certainly expose it for advertisers eventually, the same way AI Overview reporting is warming up for ads. Until then, track being named, tie it to pipeline, and refine the prompt list as you learn. The full playbook for earning that mention is in how to rank in AI Overviews.

FAQ

How do you do keyword research for B2B SaaS?

Start at the bottom of the funnel. Identify your money keywords (the commercial-intent terms buyers search at the buying moment), cut them to a shortlist of about 50 by revenue impact, read the SERP and audit your existing pages for each, then map the gaps into a content plan of refreshes (retargets) and net-new pages. Pull the data with Ahrefs or Semrush, do the clustering and prioritization in Claude, and verify the live results page by hand.

What are money keywords?

Money keywords are bottom-of-funnel, commercial-intent search terms that a buyer runs when they are close to purchasing: category queries ending in software, platform, provider, or solution; “X vs Y” comparisons; and “[competitor] alternatives” searches. They convert far better than top-of-funnel informational terms, and they usually carry a high cost-per-click, which is the market confirming they are worth competing for.

Can I use ChatGPT for keyword research?

Yes, as the judgment layer, not the data source. ChatGPT and Claude have no live volume, difficulty, or ranking data, so they cannot replace a tool like Ahrefs or Semrush, but they are excellent at clustering a real export into topics, generating long-tail and question variants, classifying intent, and turning a raw list into a prioritized plan. Pair the model with a real data source, ideally over an MCP connection.

What are long-tail keywords?

Long-tail keywords are longer, more specific search phrases with lower volume but clearer intent, like “case management software for small law firms” rather than “case management software.” They convert better because the specificity signals where the buyer is in the journey, and they matter more in the AI era because they match the sub-questions Google’s query fan-out generates and the prompts buyers type into LLMs.

How is keyword research changing with AI?

It is shifting from volume-hunting to intent-and-cluster mapping, and from keyword tracking to prompt tracking. AI Overviews absorb the clicks on broad informational terms, so research weights toward bottom-of-funnel money keywords that still convert. And because buyers now ask AI tools directly, you also track whether your brand is named in those answers, using tools like Scrunch or Profound, not just where you rank.

What are striking-distance keywords?

Striking-distance keywords are terms your site already ranks for at roughly positions 5 to 20, usually with high impressions and low clicks. They are the fastest wins available, because you are one round of improvement away from page one rather than starting from nothing. On a site with existing rankings, finding and improving these, especially the money keywords among them, is often higher-ROI than chasing brand-new terms.

How do you find keywords for SaaS SEO?

Start from the shapes buyers actually type, not a seed term dropped into a tool. For SaaS the money keywords cluster into a few patterns: category terms ending in software, platform, provider, or solution; “X vs Y” comparisons; and “[competitor] alternatives” searches. So you build the list by mapping those patterns across your niche first, then expand each one from the live SERP, where the People Also Ask box and related searches hand you the long-tail and question variants people really use. Ahrefs or Semrush fill in the volume, difficulty, and CPC, and Claude clusters and de-dupes the raw export. The find step is pattern-first and SERP-mined, not volume-first: you are looking for buying-moment terms, not the biggest number in the tool.