The first time I pulled competitor keywords, I exported over 5,000 and my head was spinning. Write all of them? Impossible, not enough manpower. Pick a few that look nice and write those? The articles come out identical to the competitors’ — why would you outrank them?
Later I broke the process into a four-layer funnel, and those 5,000 keywords ended up filtered down to 50 that could actually win — completely different efficiency. Each of these filtered keywords is a target that can rank in the top three. This article breaks down that funnel. Run through it once, and you can turn your pile of keywords into a writing list you can schedule directly.
The core is four sentences: from competitor keyword library to writable keywords, you have to pass through at least four gates — cleaning, evaluation, intent judgment, and prioritization — each one cutting volume while keeping quality; don’t blindly trust tool search volume numbers, they measure total market volume, not the effective traffic you can capture — when in doubt, go back to GSC and check real impressions; informational keywords usually make up the largest share, but the closest to money are commercial investigation and transactional types, so budget should tilt toward those two; run the funnel every quarter, write the standards into documentation, so output doesn’t depend on any one person.
Figure: Four-layer funnel from competitor keyword library to writing list (compiled by 运营GO)
Layer 1: Clean, cut in half
The pile of keywords exported from competitor tools (Ahrefs, Semrush, 5118 — any works) first gets four categories removed. This step requires no thinking, pure manual labor, but it directly cuts the volume in half, making the evaluation pressure much lighter later.
- Brand keywords: the competitor’s own brand names and product line names — people searching these are going for them specifically, writing about them is wasted effort, and you might even get flagged for piggybacking on traffic.
- No-intent keywords: too short and too broad, like industry-wide terms such as “seo” — small-to-medium sites like ours can’t rank for these short-term, forcing it is just burning money.
- Duplicate keywords: keep only one variant of the same intent, like “seo optimization,” “how to do seo,” “how to optimize seo” merged into one article — otherwise you’ll step into the keyword cannibalization trap later (this article on the site explains how to fix it).
- Irrelevant keywords: clearly outside our business scope, like a SaaS seller getting a bunch of “seo training courses” — delete directly.
After this layer, 5,000 basically leaves 2,500-3,000. There’s only one judgment standard: is the person searching this keyword our potential user? If not, delete. Don’t feel bad about “the search volume seems decent” — no matter how big the volume, if it can’t convert, it’s empty.
Layer 2: Evaluate, score on three metrics
For the remaining keywords, score each on three metrics: search volume, difficulty, relevance. Each item 1-5 points, total 3-15. I’m used to pulling a formula column in Excel or a spreadsheet to auto-calculate — don’t do it by hand, by the twentieth one you’ll be confused.
| Metric | 5-point standard | 1-point standard |
|---|---|---|
| Search volume | 1,000+ monthly searches | Under 50 monthly searches |
| Difficulty | No big-site monopoly on homepage | Homepage all authority sites |
| Relevance | Highly aligned with business | Tangentially related but hard to convert |
Don’t just look at tool numbers for search volume — tools calculate global or national totals, not effective traffic. Two practical cross-validation actions: go to GSC and check if this keyword has historical impression data (impressions mean Google already recognizes your site as relevant to this keyword); look at whether there are small sites in the top 10 search results — if small sites can rank there, it means we have a shot at this position, and the difficulty score can be pushed higher.
There’s also a lazy way to judge difficulty: search the keyword and see how many of the top 10 have “site authority” similar to yours. If there are 3 or more, this keyword is 4-5 difficulty for you; if the top 10 are all Quora, official sites, and industry giants, directly 1 point, don’t touch it.
Layer 3: Intent judgment, decide whether to do it
Search intent determines content format and conversion path. For the same batch of keywords, different intent means completely different writing — that’s also why I emphasized earlier that when merging, you have to look at intent, not just literal similarity. Intent falls into three categories:
- Informational: users want to learn something, like “what is a long-tail keyword.” Good for traffic-driving content, don’t expect direct conversion, but it builds trust — later catch them with email or on-site recommendations.
- Commercial investigation: users are comparing prices, reading reviews, like “which seo tool is better.” This is the golden position, good for comparison articles and product pages, closest to conversion.
- Transactional: users are ready to buy, like “seo tool discount code.” Directly make a landing page, search term and page title highly aligned, don’t beat around the bush.
Informational keywords usually make up the largest share — don’t do all informational right out the gate. Reserve some budget for commercial and transactional types, they’re closer to money. A practical ratio reference: informational 50%, commercial 30%, transactional 20% — small-to-medium sites don’t need to pursue full coverage, spend money where it can convert. To systematically map intent to content, reference the keyword intent mapping funnel article.
Layer 4: Prioritize, by ROI
Finally sort by “score × business value.” Business value isn’t pulled out of thin air: can the traffic this keyword brings be caught (do we have corresponding products/content), can the team sustain output (does it need long-term maintenance), after ranking will competition immediately push you down.
After the first three layers, 50 keywords sorted into three tiers: tier 1, 10 keywords to write immediately (high score, clear business value); tier 2, 20 keywords scheduled into next month’s calendar (medium score, can wait); tier 3, 20 keywords into the reserve pool (low score but might be useful later, like when brand keywords pick up then come back to do it). Looking back half a year later, if even half of these 50 keywords make it to the top three pages, the effect is much stronger than blindly spreading volume — because every article was calculated, not luck-based.
Make the process repeatable
Don’t run this funnel just once. Run it every quarter, because competitors change, keyword libraries change, and our own strength changes too. Write the keyword filtering standards into documentation, so new colleagues can run through it and get similar conclusions, instead of relying on one person’s gut every time — when the process is stable, output is stable. The documentation should fix at least three things: cleaning removal rules, three-metric scoring table, and three-tier priority thresholds. Next time someone new takes over, just fill in the numbers according to the table.
Implementation is five steps: export your current largest competitor keyword library, first run layer 1 cleaning, delete brand keywords, no-intent keywords, duplicate keywords, and irrelevant keywords cleanly; build a scoring table for the remaining keywords, search volume, difficulty, relevance each 1-5 points, difficulty score must be cross-validated in GSC; divide keywords by intent into informational, commercial, transactional three types, check whether budget allocation tilts toward commercial and transactional; sort out three priority tiers, tier 1 enters the writing calendar this week; write these standards into team documentation, set a quarterly rerun reminder.
