Competitor analysis, the working version

The competitor analysis and traffic checklist

A competitor audit goes well beyond a Semrush export: it is a structured way to read another company’s playbook and decide what is worth copying, what is worth ignoring, and what they have yet to work out. The question underneath every one we run is the same: where is their traffic actually coming from, and is that source repeatable in your language, your market and your budget?

Save this part

The competitor analysis checklist itself

Nine steps, in the order the audit runs. The sections below cover why each one is there and how to get it right.

  1. Build the real rival list: direct, indirect, content and AI-citation competitors, beyond the names on the sales whiteboard.
  2. Establish the margin of error before quoting a single traffic number to anyone.
  3. Map traffic sources and their shares: organic, paid, direct, referral and geographic mix.
  4. Read what each channel discloses about how their marketing function is actually built.
  5. Read their top pages like an editor: audience, intent and next action on every money page.
  6. Find the terms they rank for that you have still to win, and the gaps they have left open.
  7. Audit AI citations as well as SERP position.
  8. Check the multilingual setup: hreflang, localized schema, language-specific content.
  9. Cut the findings to six lines: three quick fixes, two content gaps, one structural play.

Before you quote a number

What the traffic tools tell you, and how far to trust them

Treat Semrush traffic estimates as estimates, kept separate from Analytics data when they go to a client. Published testing puts Semrush within roughly 10 to 15 percent of real traffic on mid to high traffic sites, and at 20 to 40 percent variance once a site is small. Similarweb is strongest on large international domains, and its variance widens faster on smaller ones: 15 to 30 percent reported for mid-sized sites and above 50 percent at the low end.

A 50 percent margin of error still leaves the data useful, as long as the numbers are read as direction. You can see which rival is growing, which is flat, and which mix is tilting towards paid. Quote a range: a rival estimated at 47,000 visits a month could really sit anywhere between 22,000 and 70,000, and the range is what goes to the client.

For an international audit the two tools are better together than either alone. Similarweb’s panel is stronger outside North America, so a rival earning most of its traffic in Germany or Spain is read more accurately by layering Similarweb estimates over Semrush keyword exports.

Accuracy figures from Traffic Masters’ traffic checker comparison and Bright SEO Tools on Semrush against Similarweb.

Phase one

Four kinds of competitor, beyond the one on the whiteboard

Every client hands over three or four main competitors at kickoff, and in our experience at least one of them competes offline only.

Direct rivals

Selling the same service to the same audience in the same language. The list the client already has, and usually the shortest of the four.

Indirect rivals

Solving the same problem with a different model, such as an AI tool taking on work that once went to a translation agency.

Content competitors

Publishers and blogs outranking you on your own money terms while selling something else entirely. They win the click and leave the deal for a seller.

AI citation rivals

The domains an AI assistant reaches for when someone asks an industry question. Newest of the four, and the one most worth adding to the list.

Build the real list from an organic competitors report in Ahrefs or Semrush, cross-checked against the places your buyers actually compare vendors. A pass with a handful of SEO browser extensions is usually enough to see how each one is built.

Phase two

What a channel mix discloses

The mix feeding total visits is where the insight sits: it describes what their marketing function is, and where it is thinnest.

High direct share

Real brand equity and a repeat audience, usually with offline or PR spend behind it.

High organic, low paid

A mature content programme, often with dedicated search people on staff.

High paid, low organic

An acquisition shop, whose position moves with every change in ad costs.

High referral, low search

Partnerships, affiliates, or a single viral page holding the number up.

Concentrated geography

A local focus, often with hreflang and localized content still to build.

Phase three

Read their content like an editor

Open their best pages and ask three questions. Who is it written for? What does it ask the reader to do next? Does it answer the query in full?

The pattern we meet most often is the everything hub: a few hundred articles, a good share of them short and machine-written, and traffic on a long plateau. It looks like a content engine from the outside, and every quality update opens ground for a rival with deeper pages.

Look at where their intent coverage stops. A rival that owns the how-to queries and stops short of pricing, comparisons or alternatives has left the bottom of the funnel open for you. Three sharp comparison pages, linked properly from the pages that already rank, quietly win those enquiries.

Phase four

Audit the citations as well as the position

A growing share of research now happens inside an assistant before anyone clicks anything, so a rival can sit outside the ten blue links and still be quoted in every answer. Tools track this now, Ahrefs Brand Radar among them, and their counts are worth treating as direction. Run the same prompts yourself once a month and see who gets named. Four structural patterns make a page easy to lift:

Clean data tables

Tabular content gets lifted intact, which is why a table often outranks the paragraph that says the same thing.

Question-and-answer blocks

Direct answers to specific questions, set out so a machine can lift the answer directly.

Entity linking

Their brand tied to recognised industry concepts, with outbound references to sources worth citing.

Visible expertise

Named authors, real dates, actual citations. The signals Google's own guidelines reward, and the ones LLMs reuse.

On international queries, check how they handle technical SEO across languages. A clean multilingual setup is a moat, and an untidy one is an opening for whoever takes it seriously first.

Phase five

Cut eighty findings down to six

Where an audit turns into action. Eighty issues make a list; six make a plan. Pick six: three quick technical fixes, two content gaps where a rival ranks on a weak page, and one structural play worth a quarter or two.

High

Fix technical errors and schema gaps on the money pages.

Cleaner crawling and indexing, with impressions landing on pages that can convert.

Medium

Build or rewrite content where a rival ranks on a weak page.

Organic capture on competitive queries, built over months.

Low

Tidy referral sources and link properly from the journal to the service pages.

Stronger topical authority and broader brand signals.

Before you buy anything

A first pass runs on free tools

Most of the checklist can be worked through on free tiers before anyone signs up to a paid stack. Ahrefs Webmaster Tools and the free tiers of Semrush and Ubersuggest will surface organic competitors. The Similarweb browser extension gives a quick directional read on any domain’s channel mix. Keyword Planner and your own Search Console show the queries you already touch, which is how you spot the queries a rival owns that are still open to you.

Upgrade once the exercise has proved it is worth paying for. The discipline in the checklist matters more than the subscription behind it.

Keep reading

More from the journal

See everything in SEO fundamentals