SEO competition research: how to decide whether a query is worth going after

SEO competition research is the work of deciding, before you spend anything, whether your site can reach the top of a given query. It has two layers and both are needed. The first is the difficulty score a keyword tool gives you, useful for sorting a thousand phrases into a rough order. The second is a manual read of the results page, which makes the decision: what kind of page is ranking, who holds the ten positions, how much of the page is left for an organic result, and whether the weakest result is weak for a reason you can beat. The score sorts the list. The read decides the work.

This page covers assessing the competitiveness and winnability of individual queries. For working out who your competitors are and tracking what they do, see competitor analysis and monitoring competitors in search.

What a keyword difficulty score is measuring

Every major platform publishes its own difficulty metric, and each one computes it differently from its own index. The common core is the strength of the pages currently ranking, usually derived from how many distinct sites link to those URLs, expressed on a zero to one hundred scale. Three facts follow from that, and they are the difference between using the number well and being misled by it.

• Scores are not comparable between tools. A 34 in one platform and a 34 in another are different calculations on different link indexes. Pick one platform for a project and stay in it, so your numbers are internally consistent.
• The score describes links, not content quality or intent. A query where ten thin pages rank because nobody has written a good one can score high on links and be easy in practice. The opposite happens too.
• The score is relative to nothing until you calibrate it. A difficulty of 40 means something completely different for a two year old site than for an established one.

So calibrate first. Export the queries where your site already ranks in the top ten, look up the difficulty score of each, and take the upper end of that range. That is your working ceiling. Targets at or below it are candidates now; targets well above it need links and time first. Redo it quarterly, because the ceiling moves as the site strengthens.

The seven point read of the results page

Open the query in a private window, with the location set to the market you sell in, and work down this list. Ten minutes per query, and it is the highest value ten minutes in keyword research.

• 1. Intent and page type. What kind of page is ranking: a product page, a category, a comparison, a how to guide, a tool, a forum thread, a directory. If all ten are one type and you plan a different type, the query is won by making that type, not by writing better.
• 2. Who holds the ten positions. Count them by category: national brands, publishers, marketplaces, directories, forums, and sites your own size. Two or three sites your own size is the most encouraging signal on the page.
• 3. How much room is left. Count what sits above the first organic result: ads, an AI answer, a local pack, shopping units, a video carousel, people also ask. A query where the first organic link is most of a screen down is worth less than its volume suggests.
• 4. The weakest result and why it is there. If it is a thin page on a very strong domain, you are competing against the domain. If it is a thin page on an ordinary site, that is an opening.
• 5. Depth and freshness of the ranking pages. Read the top three properly. Note length, structure, sub question coverage, original data or tools, and when they were last updated. Your specification is all of that, plus the thing none of them did.
• 6. Links at the page level. Check referring domains to the ranking URLs, not to their home pages. Pages ranking on domain strength with no links of their own are a different fight from pages carrying fifty referring domains each.
• 7. Whether it should be a page at all. If the query is a sub question and the ranking pages are all broader pages with a section on it, the right move is a section in your own broader page.

A scorecard that ends in a decision

Score each query out of ten, two points for each of the first five criteria, and write the decision in the row. The point of a scorecard is that it forces a verdict instead of a note.

Criterion	2 points	1 point	0 points
Intent match	We can make exactly the page type that ranks	We can make it with effort we are willing to spend	It needs a product or tool we do not have
Field strength	Three or more sites our size in the top ten	One or two	Ten national brands, marketplaces or publishers
Room on the page	Organic results near the top	Some features above them	Ads, an AI answer and a pack before the first link
Beatable gap	A clear content or data gap we can fill	We can match the best page	The best page is excellent and current
Link requirement	Ranking URLs have few links of their own	Moderate, and we have a plan to earn some	Heavy page level link profiles throughout

Eight or more is a go. Five to seven queues behind the eights. Four or less is a skip, written down with the reason so nobody reopens it in three months.

Your own data outranks any difficulty score

Before researching anything external, mine what the site already does. In Google Search Console, the Performance report gives you measured impressions, clicks and average position for your own queries. Two cuts of it are worth more than a difficulty column.

• Striking distance queries. Filter to average position between five and twenty with meaningful impressions. Each one is proof that Google already considers your site a candidate, so the competition question is answered and improving the page is cheaper than winning a new query.
• Queries with impressions and almost no clicks. Either the title and description are not earning the click, or you rank for something the page does not serve. Both are quick fixes.

For finding queries competitors hold and you have nothing for, a keyword gap report is the right instrument, covered in how to run a keyword gap. Treat its output as a candidate list that still has to go through the seven point read.

The shortlist document

One row per query, seven columns, and nothing else. Anything wider becomes a research project that never produces a brief.

• Query, and the monthly volume you are working from.
• Difficulty score from your chosen platform, and the page type currently ranking.
• Scorecard total out of ten.
• The gap you intend to exploit, in one sentence, and whether an existing page can carry it.
• Decision: go, queue, or skip with the reason.

What is a good keyword difficulty to target?

The one at or just below the ceiling you calculated from your own existing top ten rankings. That is a real number for your site, and it moves up as the site earns links, whereas a general rule of thumb ignores everything that matters.

Can I do competition research without a paid tool?

Yes, and the manual read is the better half of the method anyway. Search Console gives you your own measured data, the results page itself gives you the field, and the ranking pages give you the specification. A paid platform saves time on scoring a long list and on page level link counts.

How do I handle a results page dominated by directories and marketplaces?

Do both jobs. Get listed well on the directories that rank, because that is where the buyer is, and build your own page for the same query in parallel. The listing earns traffic now and the page compounds.

Should I target a query where a competitor ranks with a weak page?

That is the best kind of target, provided the page is weak on an ordinary site rather than weak on a very strong one. Check the referring domains to that specific URL before you commit, because a weak page held up by a strong domain is still a hard fight.

How often should this be redone?

Recalibrate your ceiling quarterly and re read the results page for any query you have been working on for a quarter without movement. Results pages change shape, and a query that was open when you briefed it can have a new page type by the time you publish.

The competitive field that this method does not show you

Everything above reads the ten blue links. A growing share of buyers now ask ChatGPT, Gemini, Perplexity, Claude or Google's AI Overviews first, read an answer that names two or three companies, and treat those names as the shortlist. The companies named there are frequently not the ones ranking on the same query, and the five name different companies and cite different sources as each other. Add it to the quarterly round: ask your most valuable queries in two or three assistants, leave your own brand name out of the prompt so the answer is not steered, and record who is named and what is cited. AI Knows Us runs that comparison across the five and reports the sources behind each answer, and the first scan needs no account.