SEO competitors compared, how to build the set yourself because no tool hands it to you

Every competitor tool in this market analyses a competitor set you supply. None of them decides who belongs in it, and the vendors are honest about that if you read their pages. So the set is your job, and it is the part that decides whether any of the analysis is worth reading.

What the tools actually claim, read today

We read Semrush's competitor analysis page at semrush.com/features/competitor-analysis on 21 October 2026, because it is the most used product in this category and because it is unusually plain about its own data.

On scale it publishes "808M+ domain profiles", "1B+ daily global clickstream events" and "36M+ unique brands for AI visibility analysis". On where the numbers come from, it says traffic data "comes from real clickstream data, anonymized records of over 200M internet users' online journeys", "sourced from a large panel of third-party providers updated every one to two days", then "cleaned with a proprietary machine learning model that filters anomalies".

And here is the thing worth noticing, which is an absence rather than a claim. The page does not say it identifies your competitors for you. What it offers is automatic monitoring of rivals you have already named: "Monitor competitor content, new pages, and ad activity automatically". That is a different and more modest promise than the one most buyers hear, and it is the correct promise, because a tool cannot know which companies a buyer in your market actually considers.

So the data is enormous and genuine. The judgment about who to point it at is not included.

How to build the set, in four passes

Pass one, the query competitors. Take the ten queries you most want to win. Open each one and write down who holds the top five. Count how often each domain appears. This is your search competitive set and it is frequently nothing like your commercial one. You will usually find two or three domains you had never heard of, and at least one publisher or comparison site that is not a competitor at all but is taking the traffic.

Pass two, the commercial competitors. Ask your sales team, or your own memory, who you actually lose deals to. Write that list separately. The overlap between pass one and pass two is often small, and that gap is the most useful thing on the whole page. A company that beats you in search but never in a deal is a traffic problem. A company that beats you in deals but is invisible in search is a company you can take search from.

Pass three, the substitutes. Who do people buy instead of buying anything in your category at all. Spreadsheets, doing nothing, an in house hire. These never appear in a tool and they are often the real competition.

Pass four, cut the list to five. A competitive set of twenty is a set of none, because you will not act on twenty. Keep the ones you both lose to commercially and meet in the results, and track those.

What to actually compare, once the set is right

Three things, in order of how actionable they are.

Which queries they hold that you do not have a page for at all. This is the only competitor output that converts directly into work. It is a list of pages to write.

Where they are ranking with a page that is worse than one you could write. Open the page. Read it. If it is thin and it is winning, that is authority and not quality, and you need links rather than words.

What they publish that you cannot match. Original data, a tool, a free calculator, a genuine dataset. This is the part that is hard to copy and it is the right thing to copy anyway.

And one thing to read with care. Every traffic number in every competitor report is modelled, not measured. Semrush's own description above says exactly that. Use those numbers to rank competitors against each other, never to forecast what you would get.

Your other competitor set, which is a different list

Here is the finding that changes this exercise, and it is checkable in ten minutes with no tool at all.

Ask ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews the buying question your customers ask, in their words, without naming your own business. Write down the names that come back.

That list is usually not the same as the list from pass one. An assistant naming two or three suppliers is not reading the top of a results page and reciting it. It is answering from what it has absorbed about who is credible in your category, and that frequently promotes a company with better coverage in reviews, directories and editorial over a company with better rankings. Which means you can be winning the search competitive set and losing the naming one, and no SEO report will tell you, because there is no position number in an answer with no results page.

AI Knows Us measures that second list and nothing else. We are not a competitor research tool. We do not report backlinks, keyword gaps or modelled traffic, and we are not an alternative to Semrush or Ahrefs for any of the four passes above. What we do is run those buying questions across all five surfaces on a schedule and report which names appear and which do not.

On the criterion of what a vendor publishes next to its price that a buyer can verify before paying, applied to us too: our ladder at aiknowsus.com/llms.txt, read 21 October 2026, publishes Growth at $299 a month with 100 buyer questions watched, 1,600 credits and all five engines. Where Semrush beats us outright, and it is not close, is the scale and breadth of Google side data: 808 million domain profiles and a billion daily clickstream events are not numbers we have or claim. Those are two different purchases and a serious team shopping this category should probably own both.

And our own floor, stated against ourselves: SEO Start at $149 publishes 0 buyer questions, 0 credits and 0 engines. It does not read answer engines.

What to read next

For the tools themselves, and which numbers in a competitor report are estimates, see best SEO competitor analysis and competitor research tools compared.

If the specific job is the link gap, what a link gap report can and cannot show is at best competitor backlink checkers compared. We link to it deliberately. When we measured inbound links across 2,286 stored page bodies on 21 October 2026, no other page on this site links to that one, and 939 of our 2,200 live pages were in the same position, 42.7 percent, an upper bound.