Competitor analysis in search: what to look at, and in what order

"Go and look at what the competitors are doing" sounds like a clear task right up to the moment you open the first tab. Then it turns out there are forty of them, half are marketplaces, and it is unclear what you are meant to be looking at. What follows is an order that turns this into two days of finite work ending in a concrete task list.

Who counts as a competitor

There are three different sets, and confusing them ruins the whole analysis.

• Business competitors sell the same thing to the same people. This is the list a client will give you.
• Search competitors rank beside you on your target queries. That can include trade press, a directory, or a marketplace that sells nothing directly.
• Reproducible competitors are a subset of the second, with everything struck out whose structure and link mass you could not recreate.

The third set is the one to work with. The first tells you about the business, the second builds the list, and the third turns it into tasks.

A useful side effect: it often turns out the company a client names as their main rival has no organic presence at all. Showing that in numbers at the first meeting is cheaper than showing it after a quarter of work.

Step 1. Build the list from the results

The list comes from keyword overlap, not from memory. The tool returns domains ranking for the same queries you do, sorted by how large the overlap is.

Two settings change the output more than anything else: region and search engine. For a regional project the competitor list in one city differs from another, and that is not measurement error. If the project works in both engines, the lists diverge too, and both need working through.

A normal list size at this stage is twenty to fifty domains.

Step 2. Strike out what you cannot reproduce

Twenty becomes four to six. Cross out:

• Marketplaces and aggregators. Different page type, different link mass, nothing to copy.
• Business directories and listings sites.
• Trade press, unless you are trade press.
• Sites living mainly on paid search. Check the share of paid traffic: a competitor at ninety per cent paid is not a rival in organic, even sitting beside you.

The four to six that remain are the subject of the analysis. Every step after this runs against those, not against the original forty.

Step 3. Read the page type

For each target cluster, open the top ten and look at what is there: product pages, listings, articles, roundups, video. The engine has already shown which format it treats as a valid answer to that query.

This is the most underrated part of the analysis. If ten listings hold the top ten and you are planning an article, the question is not how good the article is. That page has to be dragged uphill against the format, and it costs several times more.

While you are there, note the rich blocks: featured answers, images, video, maps. They take up space and change the arithmetic of clicks even from first position.

Step 4. The keyword gap

Simple arithmetic: take the queries your competitors rank for, subtract the ones you rank for, and look at what is left.

The result splits into three parts:

1. Queries every competitor ranks for and you do not. These are holes in your structure: everyone has a page on the topic and you do not. The cheapest growth on the list.
2. Queries only one competitor ranks for. Either they found a niche or they are wasting effort. Check the volume and decide.
3. Queries you rank for and they do not. This is your advantage, and it is worth defending rather than forgetting in favour of catching up.

The third part is almost always missing from reports, because competitor analysis gets done in one direction out of habit.

Step 5. The link gap

The same subtraction, applied to linking domains. Look for sites linking to two or more competitors and not to you: if three competitors have a link and you have none, that site is willing in principle to link to companies like yours.

The quality of the answer depends directly on how complete the tool's index is. The difference between platforms is noticeable here, and it is covered in our Semrush and Ahrefs comparison.

What to do with the list is a separate conversation, and rarely a quick one. The list is useful even without acting on it, because it shows what class of site mentions your niche at all.

Step 6. What they have that you do not

The last step is done by hand and produces the most work for the coming quarter. Open the three or four competitor pages earning them the most traffic and compare them with your equivalents point by point:

• What their page structure has that yours lacks: a specification table, a questions block, reviews, a price calculator.
• How much more thoroughly the topic is covered, and in which part.
• What the heading and the first paragraph say.
• Where the internal links from that page go.

What comes out is a list of edits to existing pages, and those usually pay back faster than new material.

What does the work, and what it costs

The webmaster console only shows your own site. Anything about other people's requires a paid tool, and that is the only reason to pay for one.

Prices from the gbseo.ru catalogue as of 19 September 2026; "retail" is the vendor's own price for the same subscription. The full set is 4 999₽ a month against 36 700₽ at retail. The calculator on the homepage prices any subset.

Recurring mistakes

Copying a strong competitor's structure wholesale. A site in the top ten is there for a combination of reasons including age and link mass. Its structure suited its situation several years ago.

Measuring the gap in one direction only. Without the reverse cut you cannot see what you already win, and those positions are the first to be lost.

Taking a competitor's traffic estimate as fact. It is a computed figure. Comparing competitors against each other inside one tool is sound; carrying an absolute number into a plan is not.

Running the analysis once. Results reshuffle, and the competitor list will be different in six months.

How often to repeat it

A full analysis is worth doing twice a year. In between, a short quarterly check covers it: has the step 2 list changed, have new rich blocks appeared on the main clusters, has anyone you were chasing slipped.

An unscheduled analysis is warranted in two cases: a sharp drop across a group of queries, and a domain appearing in the top ten that was not there before.

Keep the results of each analysis in one dated file. After a year it shows what no monthly report can: which competitors grow steadily, which stay where they are, and which appeared and vanished. That trend is more useful than any snapshot of today, because it shows who you will still be dealing with.

FAQ

How many competitors should the analysis cover?

Four to six after the cull. More adds no new information while the time spent grows linearly.

What if every competitor is a marketplace?

Then the engine treats an aggregator listing as the right answer for those queries, and fighting for them is expensive. Look at the narrower queries in the cluster, where the format of the results is different.

Can this be done with free tools?

Step 3 is done by hand and costs nothing, and it is one of the most useful. Steps 1, 4, and 5 need data about other people's sites, which is not available free.

How do you tell whether a competitor grew on links or on content?

By which links they picked up during the growth period and on which pages. If the position change coincides with links arriving on one page, the answer is immediate.

Do Yandex and Google need separate competitor analysis?

The lists differ, especially on commercial queries. A sensible compromise is a full analysis on the project's primary engine and a short list check on the second. What else differs is in our piece on Yandex and Google.