SEO Traffic Estimator, Checked Against Real Data

A traffic estimator will tell you a competitor gets 24,000 visits a month. It will not tell you the number could be half wrong, in either direction, and nothing on the screen says so.

That gap is the whole problem with an SEO traffic estimator. These tools are useful for comparing one site with another. They are poor at the job most people hire them for, which is finding out how many people actually arrive.

So use an estimator for direction and rough scale. Check its number against the visits you can count, and drop any estimate you could not defend to a client. The last section gives you the check, and it takes about an hour.

The tools are not lying. They answer a different question from the one you asked, and the rest of this post shows which question each one answers.

What an estimator is actually counting

Most estimators never see a visit. They read which keywords a domain ranks for, look up how many people search each one every month, and apply a click curve. Take a keyword with 1,000 searches a month. If the page ranks first, the model hands the site a share of those searches as visits. If it ranks seventh, the share is smaller. Multiply, sum across every keyword the tool tracks, and you have the number on the screen. Nothing in that chain looked at a visit.

That is also why a search showing your page is not the same as a visitor. The difference between impressions and clicks is exactly what an estimate skips.

On paper that is reasonable. In 2016 the SEO agency Screaming Frog tested three estimators against real analytics and put the problem plainly: the tools estimate visibility from ranking position and keyword volume, and that does not necessarily correlate to traffic. Screaming Frog, 13 June 2016

Three things break the model.

1. The tool only knows the keywords it has indexed. Long-tail queries it never tracked do not exist in its number.
2. It assumes a rank earns the same clicks whatever the search is. Brand queries, answer boxes and map packs take clicks away from the organic result, and the model cannot see them.
3. It counts rankings, not reading. A page can sit at position four and get almost nothing, and the estimate will still credit it.

Open the tool now and look at the keyword list behind the number. Sort it by estimated traffic. If the top ten are terms a buyer would type, the model has something real to work with. If they are your competitor's brand name and three product codes, the number is a guess with a chart on it.

This is where the estimate and the count you can check part ways. The keyword list tells you what the tool believes. Your analytics tells you what happened. Keep both open, because the rest of this post is about the distance between them. If you want the manual method for a site you do not own, how to estimate a website's traffic walks through it without a paid tool.

The 2016 test still shows the shape of the error

Screaming Frog took Google Analytics organic data for 25 sites over February, March and April 2016, UK organic traffic only, and compared it with what SimilarWeb, Ahrefs and SEMrush said for the same sites. Screaming Frog, 13 June 2016 The test is old, and so are the tools as they were then, so read it as a pattern rather than a benchmark.

The pattern is the point. The biggest site in the test, a charity with 8.8m actual visits, came out at +36% in SimilarWeb, -8% in Ahrefs and -37% in SEMrush. Three tools, one site, a spread of 73 points. If you had picked the wrong tool for that site, you would have reported a number that was off by more than a third.

Small sites fared worse. The author noted that the sites under 10,000 organic visits were the least accurately estimated, with almost every estimate from every tool at least 40% wrong, either over or under. Screaming Frog, 13 June 2016

The average looks better than it is. SimilarWeb's +1% is the best number in the table, but it comes from high and low errors cancelling each other out. A site 60% too high and a site 60% too low average to zero. Read the total, read the spread, and ignore the average. If the only figure a vendor gives you is the average, ask for the per-site numbers before you believe it.

Choose the traffic estimator by the question you are asking

Most arguments about which estimator is best are really arguments about different questions. Write yours in one sentence before you open a tool.

1. Which of two competitors is bigger this month? Run both through two estimators. If they agree on the order, trust the order. Do not trust the gap.
2. Is this keyword market worth entering? Use search volume for the exact terms, summed, with the date. Volume is a count of searches. Estimated traffic is a model built on top of a count.
3. What will my site get next quarter? No estimator answers this. Your own history does, and the calibration section shows how to read it.

Worked example. A competitor shows 18,000 visits in one tool and 9,000 in another. That is a two-to-one gap, so neither level is usable. Both tools still put them above you, and that ordering is what you can act on. Pick the three keywords where both tools say they rank, check where you rank for the same three in Search Console, and make that your first task for the week. If you rank on none of them, you have learned where the competitor's visits come from, which is worth more than the level.

A 2026 test of ten estimators by SemDash lands in the same place. Its author starts with the estimator that fits the question, checks the pattern against a second tool, and goes back to search data for ground truth. SemDash, traffic estimation tools test Two tools that agree are not proof. They are a cheap way to catch the one that is obviously wrong.

Calibrate the estimator on your own site

This is a one-hour job, and it is the only way to learn how much a tool misses for sites like yours. Your Search Console clicks are the count you own.

1. Pick your domain and one window: the last 28 days, ending three days ago. Search Console data lags, so the final two or three days may be incomplete. Search Console Performance report help
2. In Search Console, open Performance, set the search type to Web and the date range to that window, and note the total clicks. That number is your actual.
3. Run the same window for your own domain in the estimator. Write its number down beside yours.
4. Work out the miss. Take the estimate, subtract the actual, and divide by the actual.
5. Decide what the tool is good for. Within 25% either way, use it to compare sites in your niche. Between 25% and 50%, use it to rank competitors against each other and nothing more. Beyond 50%, keep it for direction and never quote its level.

The thresholds are a rule of thumb, not a published standard. Set your own if the decision is expensive.

A worked example with made-up numbers: a hypothetical site has 6,200 clicks in Search Console over 28 days, and the estimator says 9,800 for the same window. The miss is (9,800 minus 6,200) divided by 6,200, which is 58%. That is past the line, so the level goes and the direction stays. If that bias holds for sites in the same niche, which you should test on a second site before you rely on it, a competitor the tool shows at 12,000 is nearer 7,600.

What a good result looks like matters as much as the number. A small miss means the tool is counting something close to what you count. A large miss is not a reason to throw the tool away. Open its keyword list for your own domain and compare it with the queries in Search Console. The queries that appear in one list and not the other tell you what the tool is blind to, and that is the part you can plan around. If the clicks look low for the rank you hold, check your click-through rate against what a good CTR looks like before you blame the estimate.

Where a small brand's estimate breaks, and what to do about it

Estimators model search, and the 2016 test measured organic search only, which is the same narrow ground covered in what organic traffic counts. A newsletter, a Reddit thread, a partner's link or a paid campaign sends visits the model will never see. If those are your main channels, the estimate of your site will read low by design, and a competitor who lives on search may look bigger than they are only in the tool.

Run three checks before you trust any number.

• Is the domain more than a year old? New domains have little history, and the tools lean on history.
• Does most of the traffic come from search? If not, the estimate misses the channel that matters.
• Is the site above 10,000 visits a month? Below that, the 2016 test found most estimates at least 40% wrong.

If you cannot tick all three, the estimate is for direction only. Write down what you would spend if the level were right, then what you would spend if it were half of that. If the decision survives the miss, go ahead. If it does not, the estimate was carrying the whole plan, and the plan needs your own numbers before it gets money.

Paid channels are the one place where the count is not a model. Disclosure: The Board is a paid placement. The Board sells positions on a public leaderboard of brands, ranked by the amount paid and by nothing else, and a listing starts at $1. Impressions on the board are modelled from rank, while clicks are counted on the outbound link, so the click figure is a count the listing made. It is still a paid channel, and it goes through the same miss test as everything else in this post: check what it sent against what you were told, and only then decide whether it earns a second month.

Start with the count you own

Today, run the 28-day check on your own domain, then on one competitor in your niche, over the same window. Write the two miss numbers side by side.

1. Open Search Console and export the Web clicks for the window. Save the total.
2. Run the estimator on your domain and on the competitor for that window. Save both numbers.
3. Compute each miss and read it against the thresholds in the calibration steps.
4. Open the keyword list for your domain in the estimator and mark the queries that Search Console does not show. Those are the ones the tool believes and your site is not getting.

If your miss is small, the tool earns a second look for competitor work. If it is large, the estimate was never a count, and your search data is the number to plan around. Either way, the list of queries the tool believes and you do not get is the first piece of work this post has given you.

For the rest of the numbers behind this, the whole traffic category is the next place to go. Do the check before you open a second estimator, because the first number you trust should be one you can count.