How do I analyze my competitors’ SEO?

Competitor analysis goes wrong in two predictable ways. It studies the wrong companies, and it produces a description rather than a decision.

The useful version answers one question: what would it take to beat them for the terms that matter, and is that worth doing? Everything below serves that.

Pick the right competitors

Your search competitors are the sites that rank for the queries you want. That is frequently not the same list as the companies you pitch against.

Search your ten most important terms and write down what actually appears. For most businesses the results include review sites, marketplaces, publishers, comparison blogs, Reddit threads and directories — none of which are commercial rivals, all of which are taking the traffic you want. Analysing only your commercial competitors means studying sites that may not rank at all while ignoring the ones actually beating you.

Pick three to five, favouring the ones that appear repeatedly across your term set. Repeat appearance is the signal: a site ranking for one of your terms may have got lucky, while a site ranking for seven has something structural you can learn from.

1. The overlap analysis

Question: where do we genuinely compete, and who is winning?

Build a simple table: your priority queries down the side, you and each competitor across the top, current position in each cell. This takes an hour and it reorders most people’s priorities immediately.

What you are looking for is not the worst cell. It is the pattern. A competitor beating you on every commercial term but absent from informational ones is running a different strategy, not a better one. A competitor beating you narrowly across the board is beatable with focused work. A competitor beating you by twenty positions everywhere is an authority problem that content will not solve.

2. The gap analysis

Question: what are they capturing that we are not even competing for?

This is the highest-value output and the one that genuinely needs a paid tool, because Search Console can only show you terms you already rank for — a survivorship problem that hides exactly the set you need.

Find terms where two or more competitors rank in the top ten and you do not rank at all. Overlap matters more than volume: a term several competitors cover independently is validated demand, whereas a term one competitor covers is often a personal hobbyhorse or a legacy page.

Then cluster the raw gaps. Two hundred missing keywords is not two hundred pages — it is usually eight to twelve topics. Our keyword gap analysis guide covers the clustering method in detail.

3. The page-level teardown

Question: what is this specific page doing that mine is not?

Take the single page beating you for your most valuable term and work through it deliberately.

• Structure. List every H2. Which of those sections do you not have? This is usually the actionable finding, and it is usually three or four specific subtopics.
• Time to answer. How many words before the core question is answered? If theirs answers in two sentences and yours takes four paragraphs, that gap matters for readers, snippets and AI extraction alike.
• Specificity. What concrete, verifiable claims does it make? Numbers, dates, named sources, original data. Vague pages lose to specific ones at every level.
• Freshness. When was it last substantively updated — not just redated?
• Backing. How many sites link to that specific URL, and are they relevant?

Resist the urge to conclude “theirs is longer, ours should be longer”. Length correlates with thoroughness rather than causing ranking, and padding a page makes it worse at everything including extraction.

4. The authority comparison

Question: is this term winnable at all?

Judge authority per query rather than site-wide. A site can be genuinely authoritative in its niche and hopeless against a national publisher on a broader term, and a single site-level score obscures that completely.

For your priority terms, look at the referring domains of the pages ranking one to five, weighted for relevance rather than counted raw. If every result has hundreds of relevant referring domains and you have twelve, no amount of writing will close that this year. That is not a reason to give up — it is a reason to target the more specific version of the term, win there, and use the accumulated authority to move up later.

The useful output of this analysis is a winnability label on each term: now, with effort, or not yet. Most keyword lists have never been through this filter, which is why most content plans contain a quiet majority of terms that were never realistic.

5. The AI shortlist check

Question: when a buyer asks an assistant, who gets named?

This is the newest analysis and the one no traditional competitor tool performs, because nothing about it happens on a results page.

Take ten real buyer questions in your category and run them through ChatGPT, Gemini and Perplexity. Record who is named, in what order, and which URLs are cited. Then compare that against your ranking table from analysis one.

The results are frequently uncomfortable and always informative. It is common to find a competitor who ranks below you on Google being named first by assistants, usually because they are described consistently across third-party sources — review sites, forums, comparison posts — that engines weight heavily. It is equally common to find the cited sources are not company sites at all but Reddit threads and independent roundups, which tells you exactly where the corroboration work needs to happen.

Turning it into a verdict

The analysis is finished when you can write four sentences:

1. Who actually competes with us in search, which is usually a different list from the one we started with.
2. Which terms are winnable this year, with the not-yet ones explicitly parked rather than silently carried.
3. What specifically the winning pages have that ours do not — named subtopics, not “better content”.
4. Where the corroboration gap is, meaning which third-party sources describe competitors and not us.

If your analysis cannot produce those four sentences, it was a description. If it can, it is a roadmap, and the next quarter writes itself.

Three traps worth avoiding

Copying tactics instead of closing gaps. The most common failure. A competitor ranks well and publishes twice a week, so the conclusion is “we should publish twice a week”. But their frequency may be irrelevant to why they rank — it might be a decade of accumulated links, or one piece of original research everyone cites. Copying the visible behaviour while missing the actual cause is how teams spend a year matching a competitor’s output and none of their results.

Treating the strongest competitor as the benchmark. If one site in your space has ten times everyone’s authority, measuring yourself against it produces despair rather than a plan. Benchmark against the sites immediately above you — the ones ranking third to eighth for your priority terms. Those are the positions you can actually take this year, and the gap to them is usually a handful of specific, closeable things.

Assuming their content is good because it ranks. Plenty of mediocre pages rank on the strength of domain authority and age. If you read a competitor’s top page and think it is thin, you may be right — and that is an opportunity rather than a contradiction. The question is not “why is theirs good” but “what would have to be true for mine to be chosen over theirs”, and sometimes the answer is genuine depth on a page where they got away with less.

What to do with a competitor you cannot beat

Sometimes the honest verdict is that a term is gone. A national publisher, a marketplace, or an incumbent with fifteen years of links will hold some queries regardless of what you publish.

Three responses work better than persistence. Go more specific: the term you can win is usually two or three qualifiers deeper, converts better, and builds toward the broader one. Go where they are absent: large incumbents cover breadth badly and rarely maintain depth on narrow topics. Go where they cannot follow: original data, a tool, primary research, a genuine point of view — things that cannot be replicated by writing harder, and that also happen to be what earns AI citations.

One discipline worth keeping: re-run the overlap and AI shortlist checks quarterly, and the full analysis annually. Competitive positions drift faster than most teams re-examine them, and a competitor set chosen eighteen months ago is often measuring a race that has already moved.

Sources

• Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan and Deshpande — “GEO: Generative Engine Optimization”, ACM SIGKDD 2024 (arXiv:2311.09735); cited sources and statistics as top-performing optimisation methods.
• Google Search Console — the only first-party source of your own query, impression and position data.
• SERP feature tracking, 2026 — AI Overview presence averaging ~32.5% of monitored queries.

Frequently asked questions