Open Search Console for any page that's ranking well and look at the queries report. The page was written for one keyword. It ranks for forty. A post aimed at 'cold brew coffee ratio' quietly picks up 'how strong is cold brew,' 'cold brew vs iced coffee caffeine,' and a dozen variations nobody planned. So how many keywords should you actually write for?
Here's the straight answer to how many keywords you should use: one primary keyword per page, with a small set of related terms around it. For a typical 500 to 1,000 word article, that works out to 3 to 8 keywords total. That's it. There's no official Google number, no magic density percentage, and no penalty for targeting "too few" keywords. The rest of this article shows where that number comes from, where your keywords actually belong on the page, how the math changes when you scale to a whole site, and how to build a keyword list in about half an hour.
The Short Answer: One Primary Keyword and 3-8 Total
One primary keyword per page is the rule, and 3-8 total keywords is a sensible practical ceiling for a standard article. The primary keyword is the search query you most want to rank for. Everything else on the page supports it.
This isn't my invention, and it isn't a Google rule either. It's where practitioner consensus has landed. BKA Content, a content agency that publishes its own guidance on using keywords in content, recommends 3-8 total keywords for a 500-1,000-word piece: one primary keyword, 1-3 secondary terms, and 1-4 additional supporting terms. Ahrefs suggests something very similar: one main topic plus roughly 3-5 secondary keywords per article, with the note that a genuinely comprehensive article will often rank for many more terms than you deliberately aimed at.
Notice what both of those have in common: the number is small. Beginners almost always guess the opposite direction. They assume more keywords equals more rankings, so they shoehorn ten phrases into an 800-word post and end up ranking for none of them.
Why does a small number win? Because every keyword you add to a page is a promise about what that page covers. Target "best running shoes for flat feet" and "how to choose running shoes" and "running shoe brands compared" on one page, and you're promising three different things to three different searchers. Google has to pick one page per query, and a page that's 30% about three things usually loses to a page that's 100% about one thing.
A page that fully answers one question beats a page that partially answers five.
Now, you'll see ranking reports where a single guide sits on hundreds of queries. A thorough email marketing guide picks up "email marketing strategy," "email marketing for beginners," "how to do email marketing," and dozens more that never appeared in anyone's planning doc. I used to read that as an argument for planning more keywords per page. It's the opposite. Those rankings are a byproduct of coverage. Target one intent, cover it completely, and the long tail shows up without an invitation.
Why There's No Official Number (and What Google Actually Says)
Google has never published a recommended keyword count, and as of 2026 it still hasn't. The Search Central SEO Starter Guide, updated October 9, 2026, says to think about the words a reader might search for and to avoid keyword stuffing. It gives no target count and no target word count. Nothing.
What Google does publish is a line you shouldn't cross. The spam policies for Google web search, updated October 2, 2026, define keyword stuffing as unnaturally repeating or listing terms in an attempt to manipulate rankings. The documented examples include lists of phone numbers with no real purpose, blocks of text that repeat the same city and keyword over and over, and words out of context. Violate that and you're not competing with an unclear page. You're competing against a spam filter.
So here's my actual position: there's a floor (cover the topic) and a ceiling (don't manipulate), and the space between them is enormous. Treat 3-8 as a sensible default that keeps most people inside that space, not a target to hit. Once you can look at a page and judge whether it covers its subject, stop counting. I stopped years ago, and my rankings haven't suffered for it.
The number matters far less than the intent behind each keyword. Two pages can both target eight terms. One ranks. The other doesn't. The difference is whether each term genuinely belongs to the same search intent or was bolted on to hit a quota.
Primary, Secondary, and Supporting Keywords: The Working Framework
The 3-8 total breaks into three tiers, and knowing which tier a keyword belongs to tells you exactly what to do with it.
| Tier | How many | What it is | Where it goes |
|---|---|---|---|
| Primary | 1 | The exact query you want to rank for, chosen in advance | Title, H1, URL slug, first 100 words, one H2, meta description |
| Secondary | 1-3 | Close variations and subtopics searchers also use | H2/H3 headings, body copy, image alt text |
| Supporting | 1-4 | Related terms, synonyms, entities the topic implies | Body copy naturally, no forced placement |
| Bonus (unplanned) | Often 10-100+ | Long-tail queries you never targeted | Arrives on its own when topic coverage is deep |
That last row is the one nobody tells beginners about, and it's the reason the counting question eventually dissolves. Ahrefs' own guidance, which I mentioned above, makes this point: comprehensive articles routinely rank for far more terms than the ones they were written for. I've watched small niche sites pull the majority of their impressions from queries they never consciously chose. That's not luck. It's what happens when the page is the best answer on the internet for its actual subject.
Where Each Keyword Physically Goes
Frequency trips people up. Placement is what actually separates a page that ranks from a page that merely contains its keyword, and most guides stay vague here, so let me be exact.
The primary keyword belongs in the title tag and H1, near the front if you can swing it, with the whole thing under about 60 characters so nothing truncates in the results. The slug stays short and keeps the phrase. /flat-feet-running-shoes is done. Nothing else needed.
Then the opening. Within the first 100 words, say plainly what the page is about and use the primary phrase once. That's the entire requirement. The old "if you're searching for X, you've come to the right place" dance reads like a cold email, and Google doesn't need it.
Work the primary keyword, or a very close variant, into at least one H2 somewhere it fits without straining. Put the phrase in the meta description once, written to earn a human's click rather than to please the algorithm (Google rewrites descriptions more often than not anyway). Secondary keywords live in H2s and H3s and in body copy wherever the subtopic genuinely comes up; supporting terms simply appear as they fit.
Alt text is where people go wrong. Describe the image first. If the honest description happens to include a keyword, use it. If you're writing alt text as an SEO slot, you're doing the exact thing image guidance tells you not to. Our guide to optimizing photos for the web covers the rest, compression and file formats included.
One thing I'd push back on from older SEO advice: the idea that you need the exact keyword in every one of these locations or you "waste" the page. Google's language understanding has been good enough for years to connect "best running shoes for flat feet" with "shoes for flat-footed runners." Exact-match everywhere reads like a form letter. Write for the person who arrived from search, not for the crawler that delivered them.
Keyword Density Is a Dead Metric (Here's What to Do Instead)
Forget density. There is no official keyword-density target, and chasing one is the fastest way to write badly.
The old advice said 1-2% density, meaning your keyword appears once or twice per 100 words. That heuristic dates from an era when search engines counted words like an accountant. Modern ranking systems understand synonyms, context, and topical relevance. A page can rank first while using the exact keyword phrase three times, if competing pages use it thirty times, purely because it answers the question better.
The practical replacement for density is simple: use the primary phrase where it naturally fits (you'll usually land between 3 and 10 uses in a typical article without trying), then audit for stuffing, not for absence. Signs you've crossed into stuffing territory per Google's spam policy definition: repeating the same phrase so it reads awkwardly, listing keywords in a block that adds no meaning, or inserting city names or variations that have nothing to do with the content.
If you read the paragraph out loud and it sounds like someone talking, you're fine. If it sounds like a robot reading a thesaurus, cut it.
A quick self-check I use on every draft: search the draft for the primary phrase and count the occurrences. If it's under three in a long article, I check whether the page actually states its topic clearly somewhere. If it's over fifteen, I read every instance and delete the ones that exist only for the crawler. Nineteen times out of twenty, the natural count lands between those extremes and I change nothing.
Scaling Up: How Many Keywords Per Page Across a Whole Site
Per-page thinking breaks down at the site level, and this is where the question gets genuinely interesting. A 50-page site doesn't need 50 keywords. It needs a structure.
If you assign one primary keyword per page without checking what everyone else is targeting, you get cannibalization: two or more pages competing for the same query, splitting links and relevance between them, and often both losing to a competitor who has one focused page. The classic symptom is watching your rankings for a term flip-flop between two URLs every time Google recrawls.
The fix is clustering, and here's the working method:
- Collect everything. List every keyword you can think of for your niche, plus everything from Google's autocomplete, People Also Ask boxes, and any keyword tool you use.
- Group by intent. Put keywords that a single page could satisfyingly answer into the same cluster. "How to fix a leaky faucet," "fix dripping tap," and "why is my faucet leaking" are one cluster. "Best kitchen faucets" is a different one, because that searcher wants to buy, not repair.
- Pick a head term per cluster. The highest-volume, most representative phrase becomes the primary keyword for that page.
- Map clusters to pages. One page per cluster. No two pages share a head term.
- Write, then revisit. After publishing, check Search Console. If a page ranks for a different cluster's keywords, you've mis-grouped something and should consolidate or split.
This reframes the whole question. "How many keywords should I use?" becomes "how many distinct search intents am I willing to build a page for?" For a small business site, that might be 15-30 clusters. For a content site, hundreds. The per-page count stays small either way. The per-site count is just the number of clusters times one primary keyword each.
It's also worth being realistic about which of those clusters you can actually win. Not every keyword is worth building a page for, especially on a newer domain. I've written before about what Google queries a new site can realistically rank for, and the short version is that difficulty and your site's authority matter more than the number of keywords you can dream up. Picking the right 20 beats picking 200 at random. If you're unsure how to read the odds, our guide to what keyword difficulty actually means breaks down that metric honestly, including where it misleads.
The Four Types of Keywords (and Why the Mix Matters)
When people ask how many keywords, they're often really asking which kinds. Search marketers generally sort keywords into four types by intent, and the mix on your site should reflect it.
- Informational: the searcher wants to learn. "How to descale a kettle." Best served by blog posts and guides. Usually the largest volume bucket, and usually the lowest immediate conversion.
- Navigational: the searcher wants a specific site or brand. "Spook login." You can't realistically compete for other companies' navigational terms, and you don't need to. Just make sure your own brand terms resolve cleanly.
- Commercial investigation: the searcher is comparing before buying. "Best project management tools for freelancers." Comparison posts, reviews, and listicles serve these.
- Transactional: the searcher wants to act now. "Buy standing desk 140cm." Product and service pages, with the keyword on the page that can actually take the order.
A healthy site usually skews informational for reach and transactional for revenue. If every page on your site targets transactional keywords, you'll rank for almost nothing, because those are the most contested terms on the internet. If every page is informational, you'll get traffic that never converts. The mix is the strategy.
How to Build Your Keyword List in About 30 Minutes
If you're starting from zero, here's a concrete process that costs nothing but time. I've walked clients through this on calls and it reliably produces a usable list.
- Write your seed. One sentence describing what your business or site helps people do. Be blunt. "We sell ergonomic office chairs in Austin" beats "we're passionate about workplace wellness."
- Mine autocomplete. Type the seed into Google and note every suggestion. Do the same in the search bar with letters A through Z after your core term. Ten minutes, dozens of real queries people actually type.
- Harvest People Also Ask. For your three or four most important terms, click through the PAA boxes and record every question. Each one is a potential cluster or a section within a bigger page.
- Check your own data. If the site has any history, Search Console's Performance report shows queries where you already appear. These are gold, because Google has already voted that you're relevant.
- Classify by intent. Mark each term informational, commercial, or transactional. Now you know whether it belongs on a blog post or a product page.
- Group into clusters and cut. Merge near-duplicates, assign one primary keyword per cluster, and delete anything you can't write a genuinely useful page about. A list of 12-20 honest clusters beats a list of 200 aspirational phrases.
- Prioritize. Rank clusters by a blend of relevance to your business and how achievable the ranking looks. Start at the top.
One trap to avoid at step six: don't create a separate page for every trivial variation. "How to fix a leaky faucet" and "how to fix a dripping faucet" are the same page. Google's systems treat them as one intent. Creating two pages halves the strength of both.
Rules for Keywords: The Five That Actually Matter
Most "keyword rules" floating around are folklore. These five are the ones with real consequences.
- One intent per page: a page can rank for many queries, but only ones it fully satisfies. Don't split an intent across two pages, and don't merge two intents onto one.
- Match the format to the intent: if the top ten results for your keyword are listicles, a 2,000-word essay is the wrong format. Google is telling you what searchers click.
- Use natural language, not strings: "shoes flat feet best" as a literal phrase helps nobody. Write "the best shoes for flat feet," or just describe the topic properly.
- Never manipulate: repetition, hidden text, out-of-context keyword lists, and city-name spam are documented stuffing patterns in Google's spam policies. The upside is zero and the downside is real.
- Coverage beats counting: ten well-covered subtopics will outrank one keyword repeated ten times, every time.
FAQs
How many keywords are there in total?
Effectively infinite. New queries enter Google constantly (Google has long said a large share of daily searches have never been searched before), and there's no master list you're drawing from. What exists is the set of queries your audience actually types, which for most niches is a few hundred to a few thousand meaningful variations.
What are examples of keywords?
A keyword is whatever someone types into a search box. "How many keywords" (the query that brought you here), "best budgeting app for couples," "emergency plumber Leeds," "why is my dog shaking." Notice they're not always single words. In SEO, "keyword" means a search query of any length, including full questions.
Which are the 32 keywords?
You may have seen references to "32 keywords" in the context of Python or other programming languages, where the language reserves 32 words (like if, else, while, import) that can't be used as variable names. That's a programming concept with nothing to do with SEO. Python 3 actually has 35 keywords as of recent versions, but the point stands: different topic entirely.
Should I target the same keyword with multiple pages?
No, not deliberately. If two pages target the same primary keyword, consolidate them or sharply differentiate the intent (for example, one beginner guide and one advanced guide can coexist if they genuinely serve different searchers). Watch Search Console: if two URLs both show impressions for the same query and neither ranks well, that's cannibalization, and merging them is usually the fix.
Can a secondary keyword be the primary keyword of another page?
Yes, and this is normal in a well-built site. "Email marketing for beginners" might be a secondary keyword on your pillar page about email marketing, and the primary keyword of its own dedicated beginner guide. What you avoid is the same phrase being the primary keyword on two different pages.
How many keywords should I track?
Track as many as your tool allows, but base decisions on a curated set. For most sites, 20-50 tracked primary keywords is enough to spot trends, because rankings for secondary and long-tail terms move in sympathy with the pages that own them. Tracking thousands of keywords mostly produces noise and anxiety.
How many keywords per page on a 500-word article versus a 3,000-word one?
The short-article guidance (3-8 total for 500-1,000 words) scales loosely with length, but not linearly. A 3,000-word guide might consciously target one primary plus 5-10 secondary terms spread across its subheadings, simply because it covers more subtopics. The constraint isn't word count, it's whether each term represents a subtopic the page genuinely covers. A 3,000-word page targeting 30 keywords usually means it's 30 shallow pages stapled together.
Do keywords still matter with AI search?
Yes, differently. Tools like ChatGPT and Google's AI Overviews synthesize answers from pages that clearly and authoritatively cover a topic, which makes topic coverage and clarity more valuable than exact-match phrasing. The practical adjustment is to write pages that answer questions completely and cite-worthy, rather than pages engineered around a specific phrase. Our plain-English guide to SEO analytics in 2026 covers how to measure whether that's working, including traffic that gets blended into AI Overviews and never shows up raw in your reports. If you want a refresher on what organic traffic means in Google Analytics, start there.
Where to Put Your Effort Instead of Counting
Every section above has been quietly making the same argument, so let me state it outright: the count was never the lever. One intent per page. Clusters instead of duplicates. Coverage doing the work that repetition can't. That's the whole system, which means the answer to how many keywords you should use is simply however many the topic honestly requires: 3-8 for a normal article, one primary per cluster for your site, and not a page more.
So spend the hour differently. Pull up the top three results for your primary keyword, list every subquestion they answer, and answer all of them on your page, better, in a format a skimming human can scan. Check that the first paragraph states the subject plainly, because that's what tells Google and the reader both what they've found. That process correlates with rankings far more than any keyword count ever will, and it carries a side effect worth noticing: every phrase you were tempted to count appears naturally, the same way the long tail arrived uninvited on that email marketing guide.
If producing that kind of comprehensive page consistently is the bottleneck, that's exactly the problem Spook was built for. It finds the queries your site can realistically win, drafts SEO-optimized content in your brand's voice, and publishes it to your site, so the keyword selection and coverage work happens as part of the pipeline rather than as a separate chore you keep deferring. You can see how it works at tryspook.com.
The standard to hold yourself to, going forward: every page on your site should be able to answer the question "what one query is this the best answer for?" in a single sentence. Pages that can't are the ones that need another keyword bolted on, and they're usually the ones that shouldn't exist at all.
At Spook, we've automated exactly this judgment: identifying winnable keyword clusters, writing pages that cover them thoroughly in your voice, and getting them live without you wrangling spreadsheets of keyword counts. It's built for owners and marketers who want organic growth without turning SEO into a second job. If that sounds like your situation, it's worth a look.