How to Estimate a Website's Traffic

Somewhere in a pitch you got this month, a number showed up that nobody could actually prove.

"Fifty thousand monthly visitors." "Our newsletter reaches 12,000 readers a week." The site looks decent enough. The pitch is convincing enough. You have no way to check it, and strictly speaking, neither does the person selling it to you.

Estimating a website's traffic means using a third-party tool to guess at a number the site's own owner already knows and you don't. Every tool that does this builds its guess from a different data source, and every one of them gets it wrong by a different amount, in a different direction, depending on what kind of site you point it at.

That's not a flaw you can shop your way around. It's the whole category. The move that actually works isn't finding "the accurate one." It's picking the right tool for the size of site you're checking, running more than one, and treating the result as a range you can act on instead of a number you repeat as fact.

What a Traffic Estimate Actually Measures

No outside tool can see inside a site's own Google Analytics or Search Console account. It never gets that data, full stop. What it shows you instead is a model, built from a data source the tool itself controls, pointed at whatever domain you typed in.

There are three lineages worth knowing, because each one produces a different kind of wrong.

Clickstream panels (Similarweb, Datos) buy or partner their way into browser extensions, ISPs and app SDKs that watch where a sample of real internet users actually go, then scale that sample up to estimate a whole site's traffic. Search-visibility models (Ahrefs' traffic figure) work differently. They take every keyword a domain ranks for, multiply each ranking position by an assumed click-through rate, and add it up. That method only ever measures organic search. It has no idea about your direct traffic, your email list or your paid campaigns, a distinction Ahrefs itself has published research on and one SparkToro had to correct for after the fact in its own comparison. Hybrid models (Semrush) blend clickstream sampling with a machine-learning layer on top and claim coverage across every channel.

Mixing these up is how a decent guess turns into a bad one. A "10,000 visitors" claim from a search-only model on a site that runs mostly on email and direct traffic isn't a low estimate. It's measuring a different thing entirely, and the shortfall has nothing to do with accuracy.

Find out which lineage produced a number before you trust it. That's the whole first move, and most people skip it.

How Far Off Each Method Actually Runs

This is the part most traffic-checker roundups skip entirely: the real error rate, measured against actual analytics data, not against each other.

SparkToro ran the largest public version of this test. Over two months in 2022, 1,053 marketers voluntarily connected their real Google Analytics accounts so the company could compare four providers' estimates against a full year of actual traffic, June 2020 through June 2021, across 641 cleaned sites and 7,692 site-months of data.

A correlation of 1.0 would mean an estimate moves in perfect lockstep with the real number. None of these come close. Similarweb landed within 30% of the real figure roughly 63% of the time on the largest sites in the study, which sounds fine until you remember that leaves more than a third of checks off by over 30%, sometimes far more.

A separate study from analytics firm OWOX, run against 787 anonymized sites' real Google Analytics data and last updated in April 2026, found Similarweb's deviation from the real number sat between 57% and 61% regardless of site size, while Semrush's deviation shrank to 45% specifically on sites doing 1,000,000 or more sessions a month. Below that threshold, both tools were rough in roughly equal measure.

Put the two studies together and a pattern holds. No estimator is reliable at either extreme. The worst accuracy shows up on the smallest sites, where a rounding error in the model becomes a huge percentage swing, and on the biggest ones, where even a "good" percentage is a massive absolute error. If there's a sweet spot, it sits in the middle: sites already doing five figures of monthly traffic, checked with a clickstream tool rather than a search-only one.

That gap matters in dollars, not just in theory. If a $500 sponsored slot is priced against "20,000 monthly visitors" and the real number is closer to 9,000 once you account for a typical overestimate, you didn't pay a fair rate for reach. You paid roughly double.

Estimate a Site You Don't Own, Step by Step

1. Size it up before you check anything. A brand-new blog, a mid-size niche publisher and a major outlet need different tools and different error expectations. Guess the tier from context first (domain age, how established the brand is, what it ranks for) so you know which numbers above actually apply to it.
2. Run it through two tools from different lineages. Pair a clickstream tool with a search-modeled one, for example Similarweb's free checker alongside Ahrefs' Traffic Checker, or add Semrush's free Website Checker as a third read. Two independent guesses catch a bad one that a single tool would let through unchallenged.
3. Lean toward the lower number when money is on the line. SparkToro's data shows SEMrush is "much more often over than under," and OWOX found the same overestimation pattern in both Semrush and Similarweb. If you're deciding what a placement is worth, the low end of the spread protects your budget better than the average does.
4. Flag anything with more than a 2x spread between tools as unverifiable. When two independent estimates land 3,000 and 9,000 apart on the same domain, that's not a range you can budget against. Ask for a dated screenshot of the site's own Search Console or Analytics instead of trusting either guess.
5. Watch for the specific tells of an inflated pitch: a single spike month with no repeat traffic afterward, a domain that ranks for almost nothing yet claims tens of thousands of visitors, or a headline number that jumps by 5x or more depending on which free tool produced it.

Try it on a real case before you trust the method. Pick a mid-size blog you're considering for a guest post, run it through two tools from different lineages, and write down both numbers before you read any further in that pitch email. The spread you get back is the actual answer to "how much is this worth," not the number in the subject line.

Check Your Own Site the Right Way

You don't need to guess at your own numbers. You already have the real ones, sitting in two free tools most people check once and forget about.

Google Search Console reports exact clicks and impressions from Google's own search results, no modeling involved. Open Search, then Performance, set the date range to the last 28 days, and export it. That's what Google's own product documentation confirms it publishes, and it only covers organic search, so pair it with your analytics platform for the rest. The same report splits impressions from clicks, and those two numbers tell very different stories about the same traffic if you don't read them separately.

Google Analytics 4 covers everything the search report can't: direct visits, referrals, paid campaigns, social, email.

Together, those two beat any third-party estimate of your own site by definition, because they're not estimates. If your campaigns aren't showing up cleanly by channel in that report, tag every live link before you check the numbers instead of guessing which one sent what. And before you draw any conclusion from the organic slice specifically, know what actually counts as organic traffic in that report and what quietly gets miscounted into it.

Pull this week's Search Console number right now, before the next section. Everything below is more useful once you have your own real baseline sitting next to it.

When the Number You're Being Sold Doesn't Match What You Find

Here's where this gets practical. A newsletter offers you a sponsored slot and says it reaches "10,000 subscribers, strong opens." A blog wants $400 for a guest post and claims "50,000 monthly readers." Neither number ships with proof.

Run the domain through two tools, the way the steps above lay out. Say Similarweb comes back at 9,200 and Ahrefs' organic-only figure comes back at 3,100. That gap makes sense once you remember Ahrefs is only counting search traffic; if most of that site's real traffic is direct or email, the two numbers were never going to match, and the smaller one isn't the "wrong" one.

A newsletter's actual subscriber count won't show up in either tool anyway. That number lives inside the platform it sends from, not on the open web, so ask for that screenshot directly. It's the one figure a third-party estimate can never reach on its own.

Check what backs the domain, too. Seeing who actually links to a site tells you whether a traffic claim is even plausible, since a domain claiming five figures of monthly visitors with almost no backlink profile and nothing ranking is telling you something on its own. The same due diligence belongs in front of any link building or guest post outreach you're paying to be part of, not just the placements that call themselves sponsorships.

Some placements skip the guessing game entirely by publishing the real number instead of a third-party guess. The Board, a leaderboard where rank is sorted purely by what's paid and disclosed as such on every listing, shows each brand's actual impressions and clicks in public, and the site's own traffic runs openly through Fathom's public dashboard rather than a model anyone has to take on faith. That's the standard worth holding any paid placement to: a real, checkable number, not a claim you have to run through two other tools just to sanity-check.

What to Do With the Number Once You Have It

Set your spend against the low end of the range, not the headline figure someone quoted you. If two tools spread from 3,000 to 9,000, budget as if it's 3,000 and treat anything above that as upside you didn't pay for.

Put a 30-day checkpoint on any placement priced against a traffic claim. Re-run the same two tools before you renew, and compare the new spread against the one you started with.

A shrinking spread tells you the estimate is converging on something real. A widening one tells you the site's traffic is volatile, or the pitch was never that solid to begin with. Either way, you'll know before you renew instead of after.

And when you're the one being asked to prove your own numbers, skip the third-party guess entirely. Everything else worth tracking about where your traffic actually comes from lives under Traffic. Pull your Search Console export, screenshot your Analytics dashboard, and hand over the real thing. It's faster to produce than any estimate is to argue about, and it's the one version of this whole exercise where nobody has to trust a model.