How Accurate Are Similarweb, Semrush and Other Traffic Estimates?

Similarweb is reasonably good at showing trends and comparing large sites, and much less reliable for small ones: independent comparisons with Google Analytics found errors that often reach 100% or more, worst on sites with a few thousand visits a month. Semrush and Ahrefs estimate something narrower, search traffic only, from rankings. This article sums up what the vendors and the studies actually say, why small sites break every estimator, and which free signals you can check instead.

Published: 2026-10-16 · 7 min read · Satelo

Where traffic estimates come from

No outside tool can see another site's analytics. Each one builds a model from indirect data:

• Panel and clickstream data. Browsing data from a sample of users (via extensions, apps or data partners) is scaled up to the whole internet. Similarweb describes its own method as combining many diversified data inputs and billions of digital signals.
• Keyword rankings. Semrush, Ahrefs and Ubersuggest take the keywords a site ranks for, the estimated search volume of each, and an expected click-through rate per position, then add it up. This only covers organic search.
• Direct measurement from connected sites. Some sites share their analytics with the tool, which helps calibrate the model for everyone else.

The consequence: a panel sees few visitors of a small site, and a keyword model misses traffic from newsletters, forums, direct visits and long-tail queries it does not track. Both approaches get noisier as the site gets smaller.

What Similarweb says about its own accuracy

Similarweb is fairly open about the limits. Its data accuracy article says:

• It is an estimation tool and does not expect its numbers to align exactly with your direct measurement; it expects trend alignment.
• You may get more limited results for small sites.
• If a site's last monthly snapshot is under 5,000 visits, it displays no data. Some desktop-only features need 5,000 monthly desktop visits.

Its "Not enough data" article adds that for sites with few visits, estimates may have higher variance, so it may choose not to show them. Site owners can connect Google Analytics, after which Similarweb shows the sessions as they appear in GA, publicly or privately.

So "Similarweb says my competitor gets 40,000 visits" should be read as "Similarweb's model puts them in that range." That is useful, but it is not a measurement.

What independent comparisons found

Several groups have compared estimates with real analytics data. Their methods differ, so read them as a range of evidence rather than one verdict.

Two caveats. First, these studies are of different ages, and every vendor has updated its models since. Second, some were run by companies with commercial ties to the tools or the market, and the Screaming Frog author called his test a quick, non-scientific exercise. What they agree on is the pattern, not a precise error figure: estimates are rougher for small sites, and errors in either direction are common.

Why small sites get the worst estimates

• Too few panel users. If a site gets a few thousand visits a month, the panel may record a handful of them. Scaling a handful up to a total magnifies every random fluctuation.
• Thresholds. Similarweb hides data under 5,000 visits. Other tools show a number anyway, which can look precise without being so.
• Long-tail search. Small blogs often get much of their search traffic from many rare queries. Keyword databases track popular keywords better than rare ones, so keyword-based estimates tend to miss part of it.
• Non-search channels. Newsletters, communities and direct visits are invisible to keyword models.

The practical rule: for a site that might be under about 10,000 visits a month, treat any single estimate as a rough order of magnitude at best, and look for other evidence.

How to use estimates sensibly

1. Compare like with like. Similarweb estimates all traffic; Semrush and Ahrefs estimate organic search. If they disagree, they may both be right about different things.
2. Use trends, not totals. A steady decline over 12 months in two tools is more meaningful than the absolute number in one.
3. Cross-check two or three tools. If all of them show near zero, the site probably gets little traffic. If they disagree widely, you have learned that nobody knows.
4. Ask for real data when money is involved. Before buying a sponsored post or a website, ask the owner for read-only access to their analytics or Search Console.

Our comparison of free website traffic checkers covers what each tool's free tier shows, and our guide to checking website traffic for free covers your own site.

Free signals that are easier to verify

For small sites, a few public datasets say less but are more transparent about what they measure:

None of these gives you a visit count, and that is the point: they do not pretend to. For what a Tranco position means in practice, see our explainer on Tranco rank.

How Satelo's Quality Score handles traffic

Satelo's free site check was built around this problem. It does not claim to count anyone's visits. The score, from 0 to 100, uses mostly open data, and every component shows the evidence behind it:

• Traffic evidence (the largest weight). A Tranco position is converted into a broad monthly visits band (for example, a rank between 10,001 and 100,000 maps to "100k to 1M"). Presence in CrUX, read through Google's PageSpeed Insights, counts as evidence of real visitors. For sites in the Satelo network, Similarweb and Semrush estimates are also considered. If a site is connected to Satelo, a verified figure from its own analytics or the Satelo plugin counter replaces all estimates and is labelled as verified.
• Link evidence. The best available of Open PageRank, the Majestic Million and the domain's harmonic-centrality rank in the Common Crawl web graph. A site found in none of them gets zero here and the check says so.
• Age. The earlier of the domain's registration date (RDAP or WHOIS) and its first Wayback Machine snapshot.
• Site quality multiplier. Content size and freshness plus technical basics (HTTPS, mobile viewport, response time, page weight, PageSpeed) scale the result between half and full value. It can lower a score but never creates one on its own.

The result is shown as a score with a range of plus or minus 10 points, because a precise-looking number would overstate what open data can tell you. The limits are real: a small site with no Tranco rank and no CrUX presence will show "no public traffic signals" even if it has loyal readers. That is honest, but it is not flattering, which is why connected sites can replace it with verified numbers.

What to do next

• If someone shows you "traffic by Similarweb" for a small site, ask for read-only analytics access before paying for anything.
• Use our checklist for checking a site before buying to look past traffic numbers.
• Check the open signals for any domain in one place.

For small sites, estimates break down. Satelo's index doesn't pretend to count visits: it combines open signals with the evidence behind each number, and for network sites it shows plugin-verified pageviews. Check any site free in about 30 seconds, no signup needed.

FAQ

How accurate is Similarweb?

Similarweb describes its numbers as estimates that should match trends, not exact measurements. Independent comparisons with Google Analytics found errors that often reach 100% or more, with the weakest results for small sites.

Is Similarweb accurate for small websites?

Usually not. Similarweb shows no data when a site's last monthly snapshot is under 5,000 visits, and studies by SparkToro, Omniconvert and Promodo all found its weakest estimates on small sites.

Does Similarweb overestimate or underestimate traffic?

Both happen. Omniconvert's study of 1,787 ecommerce sites found it reported about 94% more sessions than Google Analytics, and Screaming Frog's smaller test also saw mostly overestimates, but individual sites can go either way.

Which is more accurate, Similarweb or Semrush?

They measure different things. Similarweb estimates all traffic, while Semrush's traffic figure is built from organic keyword rankings. Studies disagree on which is closer, and both are weakest on small sites.

How can I see a website's real traffic?

Only with the owner's help: read-only access to their Google Analytics or Search Console, a shared analytics dashboard, or a tool page that shows data from analytics the owner connected.

Is there a free Similarweb alternative for small sites?

No free tool measures a small site's traffic directly. Open signals such as Chrome UX Report presence, Tranco rank, and Common Crawl or Majestic link data are free and transparent, though none of them counts visits.

Read next

Free Website Traffic Checkers Compared: What You Get Without PayingFree website traffic checkers compared: what Similarweb, Semrush, Ahrefs, Ubersuggest and Satelo show without paying, their limits, and how to cross-check them.

Competitor Traffic Analysis: How Many Visitors They Get and From WhereCompetitor traffic analysis without paid tools: estimate visits, find their channels, top pages, keywords and links, then turn it into a plan for your own site.

How to Get Traffic to a New Blog: What Happens in Months 1, 3 and 6How to get traffic on a new blog: what to do in months 1, 3 and 6, why search traffic starts slow, what to measure first, and mistakes that waste year one.

How to Get Blog Traffic From Pinterest in 2026: What Actually WorksHow to get blog traffic from Pinterest in 2026: business account, claiming your site, keyword-rich pins and boards, rich pins, a weekly routine and real limits.