Competitor website traffic analysis works with estimates, never with the competitor’s real numbers, and the estimates are rough. In a comparison of 184 sites against Search Console data, Ahrefs, Semrush and Similarweb missed by about half on average, and not in the same direction: Ahrefs tended to undercount and the other two to overcount. That doesn’t make the analysis useless. It means the useful readings are the ones the error cancels out of: trends over months, shares within a competitor set measured with one tool, which pages and page types carry the traffic, and how much of it is the competitor’s own brand. Read that way, a tool that is wrong about the totals is still right about most of what a competitor analysis needs.
Below: what competitor traffic analysis can and can’t tell you, where each tool’s numbers come from, how accurate they are, what AI Overviews did to the estimates, and seven steps that hold up against all of it.
What competitor traffic analysis can and can’t tell you
In a competitor website traffic analysis, the tools can show the order of magnitude of a competitor’s traffic, its direction over a year, the split between organic, paid, direct and referral, the pages and folders that receive it, the queries behind the organic part, and the countries and devices it comes from. They also report engagement and audience figures: unique visitors, bounce rate, visit duration, pages per visit, demographics, and audience overlap between two sites. Those are a model built on top of the traffic model, and they are the least accurate numbers in the product, as the pages-per-visit row in the accuracy table below shows. They can’t show the competitor’s analytics, conversions or revenue, and no tool has them. Google Analytics 4’s benchmarking compares a property with an industry average built from sites that opted in, never with a named competitor.
Some traffic is invisible to all of them. Visits from messengers, private communities and most email clicks arrive without a referrer and land in direct, so a competitor whose growth comes from a Telegram channel or a newsletter looks like a brand that people suddenly started typing. And AI assistants are a channel the traffic tools barely measure: how often a competitor is cited in ChatGPT or Perplexity answers is a separate question with its own tools, and the referral clicks it produces are small next to what the citations do for the brand.
Where competitor traffic numbers come from
The tools reach their numbers by different methods, and the method predicts the direction of the error.
| Tool and report | How the number is built | Usual direction of error | Weak spot |
|---|---|---|---|
| Ahrefs organic traffic; Semrush Organic Research | The keywords a site ranks for, each one’s search volume, and a click-through rate for its position | Under, for Ahrefs in the 184-site comparison | Only organic search; misses queries the tool doesn’t track; inherits every error in the volume estimates and the CTR curve |
| Semrush Traffic Analytics | Clickstream data from third-party providers, modelled up to all visits | Over, on 112 of the 184 sites | Small sites, where the panel holds few real visitors |
| Similarweb | Four sources by its own description: sites that share their analytics directly, its own consumer products, data partners such as ISPs and ad platforms, and public web data | Over | Sites under roughly 100,000 visits a month; pages per visit |
SE Ranking, SpyFu and Serpstat build their organic numbers the same way as the first row, from rankings, volumes and a click curve, so the same weak spots apply. The practical consequence is that numbers from different tools can’t be compared with each other. A competitor at 40,000 in Ahrefs and your site at 60,000 in Similarweb is a comparison of two methods, not of two sites.
How accurate competitor traffic estimates are
Four comparisons of competitor traffic estimates against real data are public, and they agree on the shape more than on the numbers.
| Comparison | Sample and baseline | Result |
|---|---|---|
| Collaborator / Promodo, published December 2024 | 184 sites, first half of 2024, against Google Search Console | Average error: Ahrefs 48.6%, Similarweb 57.0%, Semrush 61.6%. Ahrefs tended to undercount; Semrush overcounted on 112 of 184 sites |
| SparkToro | 641 sites, twelve months of real figures | Similarweb within 30% of real traffic for 62.9% of sites above 250,000 monthly users |
| Omniconvert | 1,787 ecommerce sites with read-only analytics access | Similarweb overreported sessions by about 94% |
| SaaS Odds live test, September 2026 | One site with public analytics, about 20,000 sessions a month | Similarweb about 88% over on visits; 3.22 pages per visit against a real 1.15 |
Three things about accuracy hold across all four. Big sites are estimated better than small ones, and under a few thousand visits a month no tool is worth quoting. The errors are systematic per tool rather than random, which is what makes the calibration in step 2 possible. And the size of the error didn’t depend on the niche in the 184-site comparison, so there’s no category where a tool can be taken at its word.
AI Overviews and the CTR curve
The ranking-based estimates have a newer problem. They turn positions into visits with a click-through curve, and on queries with an AI Overview that curve no longer describes what users do. Ahrefs’ study of 300,000 keywords, published in May 2026, found the top-ranking page getting 58% fewer clicks with an AI Overview present than without, up from 34.5% eight months earlier. Any estimate that applies a pre-AI-Overview curve overstates the organic traffic of a competitor whose rankings sit on informational queries, and overstates it most for the competitor ranking best. Two competitors with the same rankings can now have very different traffic, depending on how many of their queries carry an AI Overview, and an estimate that doesn’t know which queries do can’t tell them apart.
How to analyze competitor website traffic in seven steps
1. Build the competitor set from the SERP
The competitors that matter for a competitor website traffic analysis are the domains that rank where you want to, not the ones sales names. Take the twenty non-brand queries that bring you the most impressions in Search Console and list the domains in the top five for most of them. On most sites the set includes a publisher, a marketplace or a review platform nobody in the company thinks of as a competitor, and those are often the sites taking the clicks. It’s the same set the backlink audit guide builds for links.
2. Calibrate the tool on your own site
Because each tool’s error is systematic, your own site tells you its size. Run the tool on your domain for the same months you’ll read competitors for, and set its estimate against a baseline you trust. For organic, that is Search Console clicks summed by page, which are steadier than the query-level numbers. For total traffic, server logs are better than GA4, which undercounts wherever visitors decline cookies or block scripts. The ratio is the correction: if the tool shows your site at 1.6 times its real organic clicks, read competitors of a similar size through the same factor.
The factor holds only for sites of roughly your size, because the error grows as traffic shrinks, and only within one tool and one report. If your own data is unreliable, there are two substitutes. One is sites in your space that publish their analytics openly, as some independent and open-startup projects do on public dashboards, which is how the September 2026 test above was run. The other is the spread between three tools on the same competitor, read as the range the real number probably sits in.
3. Read trends and shares, not totals
Twelve months, one tool, one metric, the whole competitor set on one chart. A competitor that grew 40% while the rest of the set stayed flat grew, whatever the absolute numbers are, because the tool’s error is in all of them. Compare year over year, so seasonality doesn’t read as growth or decline, and filter to the country you sell in: a worldwide total for a competitor with half its visits from a market you don’t serve says little about yours. Traffic share, each site’s part of the set’s total, is the second reading and the nearest thing these tools give to market share: a competitor going from 15% to 25% of the set’s organic traffic in a year has taken something, and the remaining steps find what. On SaaS competitors, check the split by subdomain before reading any total, because app., login and docs subdomains carry logged-in customers, and a total that includes them measures the customer base rather than the marketing site.
4. Split brand from non-brand
A large share of what tools report as a competitor’s organic traffic is often people searching its name. Filter the organic keyword report by the brand and its misspellings. A competitor whose traffic grew on brand queries grew because of something outside search, a campaign, press, a product launch, and there is nothing in the SERP to take from it. Google Trends, comparing the competitors’ brand names over the same year, is a free check on that reading and doesn’t depend on any traffic model. Growth on non-brand queries is the part a content and SEO plan can compete for, and it is usually a smaller number than the headline.
5. Find the pages and page types that carry traffic
Group the competitor’s top pages by template rather than reading them as a list: the blog, the category or solution pages, the comparison pages, the tools, the pricing page. A competitor with a million visits mostly to informational blog posts and a few thousand to its commercial pages is a different threat from one whose traffic sits on pages that sell, and a domain total hides the difference. The pattern that matters is a page type the competitor has and you don’t, ranking on queries you want. Ten comparison pages bringing a competitor a fifth of its non-brand traffic is a gap with a shape; a single viral post is not.
6. Turn the gap into keywords
The keyword gap report, your domain against the set, lists the queries the competitors rank for and you don’t. Filter it to non-brand queries in the top ten for at least two competitors, and check each candidate against the live SERP before it goes on a plan, because the tools’ difficulty scores are models of their own. This query is an example: in Semrush in November 2026, “competitor website traffic analysis” shows a keyword difficulty of 4 and its rewording “competitor analysis website traffic” shows 48, for result pages that are largely the same. The five checks that reject a keyword are what to run on the list before anything gets written.
7. Read the other channels: paid and referral
Similarweb and Semrush Traffic Analytics split a site’s traffic sources into organic search, paid search, direct, referral, social, email and display. The split is more useful as proportions than as counts: a competitor at 70% organic and one at 50% paid are running different businesses, even if both totals are wrong. For paid search, the estimates of a competitor’s budget are the least reliable number in any of these tools. What can be seen directly is free: Google’s Ads Transparency Center and Meta’s Ad Library show a competitor’s live ads, the landing pages they point to, and how long each has been running. A landing page that has carried ads for months is one the competitor keeps paying to send people to, which says what it is selling hardest; it doesn’t say whether the page converts. For referral traffic, Similarweb’s referral list sorted by change shows new sources: a partner, a directory, a review site, an affiliate programme. A referral source growing for a competitor is usually a placement you can ask for too. Social gets the same reading: which network sends the visits matters more than how many the tool says there are.
What goes into a competitor website traffic analysis report
| Reading | Source | How to label it |
|---|---|---|
| Trend per competitor, 12 months | One tool, one report | Direction and percentage change; no absolute totals in the headline |
| Share of the set | Same tool | Share now and a year ago |
| Brand and non-brand | Organic keywords, brand filter | Non-brand as the competitive number |
| Page types carrying traffic | Top pages, grouped by template | The page types you lack, with the queries |
| Keyword gap, checked | Gap report plus the live SERP | The short list that passed, not the export |
| Absolute traffic, if asked | Tool estimate through the calibration factor | As a range, with the tool and month named |
Competitor traffic analysis FAQ
Can I see a competitor’s Google Analytics?
No, and no tool can. GA4’s benchmarking compares your property with an anonymous peer group of sites that opted in to share data. It never names a competitor or shows one site’s figures.
Which tool is the most accurate for competitor traffic?
It depends on the channel and the size of the site. For organic traffic, Ahrefs had the lowest average error in the 184-site comparison and erred low; for total visits on large sites, Similarweb came within 30% for most of SparkToro’s sample. For a site under a few thousand visits a month, none of them is reliable enough to quote.
How often should competitor traffic be checked?
Quarterly, with the same set and the same tool, so the readings stay comparable. Monitoring more often, or setting alerts on weekly changes, mostly tracks the noise in the estimates. The exception is a core update: a check of the set straight after the rollout shows whether the whole category moved or only some of its sites, which is the competitor-set test from the website rankings loss audit.
What I don’t know is how far the AI Overview gap has opened between the tools’ click curves and real behaviour by query type. Ahrefs published the drop for the top position, and none of the tools says whether its traffic model has been adjusted for it. Until one does, I read organic estimates for informational competitors as a ceiling rather than a figure. Building the competitor set and reading it against your own data is part of our SEO audit services.