Competitor SEO analysis: a repeatable process

Most competitor analysis starts with the wrong list. Someone opens a spreadsheet, types in the four firms they lose pitches to, audits their websites, and comes away with a document full of observations that explain nothing about why they are on page two.

The reason is simple. The businesses you compete with for customers and the pages you compete with for a search result are two different sets, and they overlap less than anyone expects.

Your search competitors are usually not your business competitors

Search does not return companies. It returns pages that a system has judged to be reasonable answers to a query. Some of those pages belong to firms like yours. Many do not.

Take an illustrative example. Marchant & Slate is a two-person commercial cleaning firm in Bristol. (This business is invented to make the point; nothing here is measured data.) They search "office cleaning bristol" and find a results page holding: a national franchise with three hundred branch pages, two business directories, a local pack of map listings, a "best office cleaners in Bristol" listicle written by a marketing site, and one actual local competitor at the bottom.

Auditing that one local competitor tells them almost nothing. Six of the seven things standing between them and the top of the page are not cleaning companies at all. They are publishers, aggregators and platforms whose entire business model is occupying results like this one. They have different economics, different content operations, and in the directories' case a structural advantage: their page can list twenty cleaners, and yours can only be one.

This reframing is the most useful part of any competitor analysis, and it changes what you do next. You are not trying to be a better cleaning company than the directory. You are trying to work out whether a single-business page can occupy that slot at all, and if so, what shape it needs to be.

It also means your competitor set is defined per query, not per business. "Office cleaning bristol" and "how often should an office be cleaned" may share no competing pages whatsoever. Analyse the query, not the industry.

Build the list by reading results, not a dashboard

The whole process runs off the results page. You need no subscription to do the core of it.

One. Write the queries first. Not the ones you wish people used — the ones they actually type, taken from your own enquiries, your call notes, your site search box and the questions customers ask before they buy. If that list is thin, building a keyword list without paying for a tool covers how to assemble one from free sources.

Two. Search in a clean context. Log out, use a private window, and make sure your location is the one you care about. Google's documentation is clear that results are localised and personalised, so the page you see logged in on your own laptop is not the page a stranger in your target city sees. If you serve a specific area, check that area explicitly rather than assuming.

Three. For every result on page one, record four things: who publishes it, what type of page it is, what the page is trying to do for its reader, and whether it is even the same kind of thing as your page.

That fourth column is the one people skip and the one that carries the information.

What the mix of results is telling you

Before you look at any individual page, look at the shape of the whole page. The composition is a statement about what the query is understood to mean.

| What dominates the results | What it usually means | Reasonable response | | --- | --- | --- | | Directories and marketplaces | The query is a shortlisting query — searchers want options, not one supplier | Get listed where the directories are; a single service page will struggle to displace a list | | A local map pack above everything | Proximity and business profile matter more than your website | Work on the business profile and local signals first; see local search work | | Big publisher listicles | The query reads as research, not purchase | Either write something genuinely better for researchers, or target a purchase-intent query instead | | Forums and community threads | People want candid, non-commercial answers | A polished sales page will not fit; consider being useful in those spaces instead | | Product or service pages from firms like yours | The query is transactional and a page like yours belongs | This is a winnable shape — now do the page-level comparison | | Documentation, government or institutional pages | The query is informational and trust-heavy | Usually not worth chasing commercially |

If your page type does not appear anywhere in the top ten, that is a finding, not a challenge. It is the cheapest insight competitor analysis produces and the one most often ignored.

Reading one competing page properly

Once you have a query where your kind of page can compete, take the two or three best results and read them as an assessor would, not as a rival.

The questions worth asking:

What does this page actually answer, and in what order? Write out its structure as a list of questions it resolves. That list is more useful than any other artefact you will produce. It shows you what the page treats as important and where it puts the thing the reader came for.

What proof does it carry? Photographs of real work, named staff, prices, licence numbers, specifications, a service area map, dates. Proof is often the whole difference between two pages covering the same topic.

Why would anyone believe this publisher on this subject? A page about roof repair on a roofing company's site and the same page on a general home-advice site are different objects. Google's guidance on helpful content asks, in effect, whether the site has a reason to be producing this. Ask it of your competitor, then ask it of yourself.

How is the page wired into the rest of the site? Which internal pages link to it, and with what anchor text? Strong competing pages are rarely orphans. This is copyable in an afternoon and costs nothing.

Who links to it from outside, and could you plausibly earn anything similar? If the links are trade-press coverage and supplier directories, that is reachable. If they are wire-service syndication from a decade of national PR, it is not. How links are actually valued and how to judge whether a given link is worth anything both matter more here than raw counts.

What does the page fail to do? This is where you find your opening.

Note what you are not measuring: word count. Google's documentation states there is no preferred or ideal word count for ranking. A competitor's 3,000-word page is not beating you because it is 3,000 words, and matching the number is not a strategy.

The gap is rarely "more words"

"Content gap" has been flattened into "they mention things we don't". Useful gaps come in several shapes, and only one of them is about coverage.

• Intent gap. The winning pages answer a different question from the one you assumed. Your page sells; theirs compares. Yours explains; theirs prices.
• Coverage gap. There is a real question every competing page leaves unanswered. Rare, and valuable when true.
• Format gap. Everyone has written prose and the reader needed a table, a calculator, a checklist or a photograph of the thing.
• Proof gap. Same claims, but one page evidences them and the others assert them.
• Access gap. The information exists but is behind a form, a phone call or a PDF. Publishing it plainly is a genuine advantage.
• Freshness gap. Prices, regulations and specifications have moved and the incumbent pages have not.

For Marchant & Slate in the illustration above, the gap on the listicle results was an access gap: every competing page invited a call for a quote, and none published so much as a rate band. A page that shows how commercial cleaning is priced, with honest ranges and the factors that move them, is a different object from a fifth "why choose us".

Is this term winnable at all

Run each target query through the same questions before spending anything on it.

1. Is any result the same kind of page as mine? If no, stop. Choose a different query.
2. How much organic room is there? Count what sits above the first organic result: ads, a map pack, shopping units, an AI summary, people-also-ask. A first-place organic result under a crowded page is worth less than a first-place result on an empty one.
3. Are the incumbents maintained or abandoned? A top result last updated years ago, with dead links and stale prices, is a softer target than one revised this quarter.
4. Do the winners hold links I could realistically earn? Not match — earn. If the answer is no, the timeline extends by years.
5. How long am I willing to wait? What the evidence supports about SEO timelines is the honest version of this, and it is measured in months at best.
6. What is one customer worth? A query you cannot win in eighteen months is still worth pursuing if a single client pays for the year. Usually it is not.

If four of the six answers are discouraging, that term is not your battle. Pick the adjacent, less contested version of it — the longer, more specific query where the results page is thinner and the intent is closer to buying.

And if the honest conclusion is that organic placement for a commercial term is out of reach on your timescale, the alternatives are paid ones: search ads, or a paid placement on a platform that already occupies the results. Our own public spend-ranked leaderboard is one such placement, and an honest comparison of paid against organic costs sets out when each makes sense. It would be dishonest to leave out the other half: what paid placement cannot do for you includes building the pages that earn rankings on their own. Buying visibility rents attention. It does not produce an asset. If you can win the organic term, win it.

What to copy and what to ignore

| Worth copying | Worth ignoring | | --- | --- | | The set of questions their page answers | Their word count | | The order they answer them in | Their design and colour scheme | | Formats they use — tables, specs, photographs | Keyword density and other counting rituals | | Types of proof they show | Third-party "authority" scores | | How they link internally to the page | Rankings they hold for their own brand name | | The objections they handle before you ask | Their ad spend, which you cannot see anyway |

The pattern is consistent: copy decisions about substance and structure, ignore surface. Anyone can rearrange a page to look like a competitor's. Almost nobody bothers to answer the six questions the competitor answers that they do not.

Tool numbers are estimates, and treating them as facts is the standard error

Third-party monthly traffic figures, keyword volumes and difficulty scores are modelled. They are built from panel and clickstream sampling, extrapolation and proprietary formulas. No third party can see another site's analytics. Difficulty scores are one company's model output, not a property of the query.

That does not make them useless. They are good at three things: surfacing pages and queries you would never have found manually, giving you a rough relative ordering, and showing change over time within one tool's own consistent methodology.

They are bad at exactly one thing, and it is the thing they are most used for: being quoted as fact. "This keyword has 2,400 searches and a difficulty of 34" reads like a measurement and is closer to an opinion. Decisions worth money should rest on the results page you can read directly.

If you want the mechanics underneath all of this, how ranking works in plain terms is the shortest honest version.

Mistakes that keep recurring

• Building the competitor list from business rivals rather than from the results page.
• Auditing the whole competing site instead of the specific page that outranks you.
• Treating a directory or marketplace as a competitor to beat rather than a channel to be listed on.
• Reading page one while logged in, in the wrong location, and drawing conclusions from personalised results.
• Chasing the head term because it feels important, when the specific version converts better and is winnable.
• Copying length instead of substance.
• Doing the analysis once and never revisiting it. Results pages change shape.

Questions people ask

How often should I redo this? Quarterly for your handful of commercial queries, and immediately if a page drops noticeably. Full analysis of everything, once a year at most.

How many competitors should I analyse per query? Two or three properly beats ten superficially. The top results usually share the pattern; you only need to see it once.

What if my competitors are clearly breaking the rules? Read Google's spam policies before assuming so — much of what looks like manipulation is not covered by them. If something genuinely violates the policies, report it and move on. Building your plan around someone else's shortcut is how you end up with their problems.

What if the top result is just a bigger brand? Then compete on specificity. Broad terms favour scale; precise, local and technical terms often do not. That is where a smaller site's advantage lives.

Is any of this worth it for a brand new site? Partly. The diagnostic value is real, but the timeline is not. A realistic timeline for a brand new site will save you more disappointment than any audit.

The most valuable outcome of a competitor analysis is often the decision not to compete — to look at a results page owned by three national directories and a franchise, and decide your effort belongs on the twelve specific queries where you can genuinely be the best answer. That is a real result. It just does not feel like one, because nobody sells a subscription for it.