HIGHER
OR LOWER

Google Trends does not show how many times anything was searched, and it has never claimed to. What it shows is a proportion, rescaled so that the highest point in whatever window you happened to choose becomes 100 and everything else is measured against that.

This matters more than it sounds, because almost every confident statement anybody makes about search volume rests on a misreading of that one design decision.

Key Takeaways

• Google Trends divides each data point "by the total searches of the geography and time range it represents", then scales the result "on a range of 0 to 100". It is a share, not a count.
• Change the date range and every number on the chart changes, because the value of 100 is reassigned to whatever the new peak is.
• Trends drops several things that readers rarely check for: low-volume terms show as 0, repeated searches by the same person in a short period are removed, and queries containing apostrophes and other special characters are filtered out.
• Absolute figures come from Google Ads instead, where "avg. monthly searches" is a 12-month average, rounded, and lumps a keyword together with its close variants.
• People misjudge frequency because they judge it by how easily examples come to mind, an effect Tversky and Kahneman named the availability heuristic in Science in 1974.

Does Google Trends show real search volume?

Not in the sense almost everyone means. It shows relative interest, and it is explicit about the arithmetic.

Google's own documentation sets out the two steps. First, "each data point is divided by the total searches of the geography and time range it represents to compare relative popularity". Second, "the resulting numbers are then scaled on a range of 0 to 100 based on a topic's proportion to all searches on all topics".

Two consequences follow, and both are routinely missed.

A term can climb on Trends while being searched less often in absolute terms, if total search traffic in that place fell faster than the term did. The line is a ratio, and ratios move when either half moves.

And the chart has no fixed vertical axis. The proportion is computed first, and only then is the largest proportion in your selected range set to 100, so widening a date range from one year to five can flatten a mountain into a foothill without a single underlying number having changed. Google's help page is blunt about the limits of the whole exercise: the data "is not a perfect mirror of search activity" and "should always be considered as one data point among others before drawing conclusions".

One thing the documentation does not spell out is what happens when you put two terms on the same chart. It describes the normalisation for a topic, not for a comparison, so the safe reading is the conservative one: because both lines are proportions of the same total, their ratio is meaningful, while the height of either line on its own is not. A term that never touches 100 in a comparison is not being reported as rare. It is being reported as smaller than the other term's best week.

What Google Trends leaves out

Quite a lot, and the exclusions are the part that catches people who are comparing two rare things.

Google lists them plainly. "Trends only shows data for popular terms, so search terms with low volume appear as 0." Anything genuinely obscure is not small on the chart; it is absent, which looks identical to zero interest and is not the same thing.

The service also "eliminates repeated searches from the same person over a short period of time", so obsessive checking by a few people does not register as popularity. It "filters out queries with apostrophes and other special characters", which quietly removes a slice of natural English. And it drops "searches made by Google products and services", a category that now expressly "includes internal searches made by AI Mode and AI Overviews".

None of that is hidden. It is simply skipped, because a chart with numbers on it looks like a measurement, and a measurement looks like it needs no reading instructions.

Google Ads is where the absolute numbers live

They come from the advertising side of the house, and they arrive with a different set of compromises.

The figures quoted as monthly search volume, including the ones behind guessing games like Asked AI More?, originate in Google Ads keyword data rather than Trends. Google describes that number as "the average number of times people have searched for a keyword and its close variants based on the month range as well as the location and Search Network settings you selected", and says that "by default, the number of searches for the term is averaged over a 12-month period".

Three things are worth holding on to there. It is an average over a year, so a term that spikes for one week in December and vanishes for the rest of the year looks modest. It bundles the keyword with its close variants, so you are seeing a family of phrasings rather than a string. And the numbers are rounded: Google warns that "your search volume statistics are rounded". Rounding is also why the volumes for individual countries or regions will not add up to the figure you see for a wider area.

Set the two sources side by side and the difference is not a matter of precision. They are answering different questions.

Property	Google Trends	Google Ads keyword data
What it reports	Share of all searches	Count of searches
Scale	Rescaled 0 to 100	An average monthly number
Time basis	Whatever range you select	12 months by default
Precision	Index, no underlying total shown	Rounded
Term matching	The term as entered	The keyword plus close variants
Rare terms	Shown as 0	Reported, subject to rounding

So even the honest absolute figure is a rounded yearly average of a cluster of related phrasings in one country. It is a good number. It is not a headcount.

Why people guess search volume badly

Because we estimate how often something happens by how easily an example comes to mind, and that is not the same thing as how often it happens.

Amos Tversky and Daniel Kahneman set this out in Science in 1974, giving it the name availability. Their illustration is domestic: "one may assess the risk of heart attack among middle-aged people by recalling such occurrences among one's acquaintances." Convenient, and biased by whoever you happen to know.

Their sharpest demonstration involves letters. Asked whether more English words begin with the letter R or have R in third position, most people say the first. They are wrong, and they are wrong in a specific, predictable way: it is easy to search memory by first letter and hard to search by third letter, so the easier category feels bigger. The paper notes people make the same error for other consonants "that are more frequent in the third position than in the first".

Search volume is a frequency question wearing a disguise, so it inherits the whole problem. Anything dramatic, recent or heavily reported is easy to recall and therefore feels heavily searched. Meanwhile the enormous, boring, constant queries, the ones about weather and banks and how to spell a word, are invisible precisely because nobody remembers making them.

There is a second twist specific to search. People type into a search box the things they do not already know. Familiarity suppresses searching, so the most universally understood subjects can be searched far less than niche ones that nobody can quite remember the details of. It is the mirror image of the problem in a guessing game like What Are You Thinking Of?, where the winning move is to work out what you cannot predict.

What happened when Google tried to read the flu off search

It failed, publicly, and the post-mortem is the best short lesson available on what search data is and is not.

Google Flu Trends estimated influenza prevalence from search behaviour. In February 2013, as Science later recounted, it "was predicting more than double the proportion of doctor visits for influenza-like illness" that the United States Centers for Disease Control and Prevention recorded. It was not a one-off: from 21 August 2011 to 1 September 2013 the system "reported overly high flu prevalence 100 out of 108 weeks".

The diagnosis in that paper, by David Lazer and colleagues, has two halves. The first is overfitting: the method was "to find the best matches among 50 million search terms to fit 1152 data points", and with that many candidates some will match by luck. The developers had already had to weed out seasonal terms that correlated beautifully with flu and had nothing to do with it, "such as those regarding high school basketball".

The second half is more interesting for anyone reading a search chart today. The authors call it algorithm dynamics: "the changes made by engineers to improve the commercial service and by consumers in using that service". Search is a product under constant revision, and the paper notes that Google's own blog "reported 86 changes in June and July 2012 alone". Their conclusion is the line to remember: "search behavior is not just exogenously determined, it is also endogenously cultivated by the service provider."

Which is the honest summary of the whole subject. Search volume measures what people typed into a box that someone else keeps redesigning. It is real data about real curiosity, and it is not a census of the world.

If you would rather test your instincts on something with a fixed answer, the quizzes hub collects the guessing games. The daily puzzles change once a day, the same for everyone.

Frequently asked questions

Does Google Trends show the number of searches?

No. Trends divides each data point by the total searches for that place and period, then rescales the result onto a range of 0 to 100. A value of 50 means half as much relative interest as the peak point in your selected range, not fifty searches or fifty thousand. Google itself describes the output as relative popularity and warns that it "is not a perfect mirror of search activity".

What does a Google Trends score of 100 mean?

It means that point is the highest in the range you are looking at, and nothing more. The value 100 is reassigned every time you change the search term, the country or the date range, so the same week can read 100 in one chart and 12 in another. That is why two Trends screenshots with different settings cannot be compared, even when they show the same term.

Why do two terms swap places when I change the date range?

Because the scale is recalculated for whatever window you asked for. Widening the range brings in a bigger peak, everything before it is squashed down against the new maximum, and the shape of the comparison changes. Nothing about the underlying searches changed; only the yardstick did.

Where can you get actual search volume numbers?

From Google Ads keyword data rather than Trends. Those figures are averaged over a 12-month period, rounded rather than exact, tied to the location and network settings selected, and they combine a keyword with its close variants. That makes them genuinely useful for comparing the size of two topics, and unsuitable for claiming a precise count of anything.

Why are people so bad at guessing what gets searched?

Because we judge how common something is by how easily we can call an example to mind, which favours the vivid and the recent over the ordinary and constant. Tversky and Kahneman demonstrated the effect in 1974 with letters: most people believe more words start with R than have R in third position, and the opposite is true. Search adds its own twist, since people mostly search for things they do not already know.

Sources

• Google Trends Help, FAQ about Google Trends data - retrieved 10 August 2026, help article at support.google.com
• Google Ads Help, About Keyword Planner search volume and forecasts - retrieved 10 August 2026, help article at support.google.com
• Amos Tversky and Daniel Kahneman, Judgment under Uncertainty: Heuristics and Biases, Science, New Series, Vol. 185, No. 4157, 27 September 1974, pages 1124 to 1131 - retrieved 10 August 2026, scan hosted by University of California, Irvine
• David Lazer, Ryan Kennedy, Gary King and Alessandro Vespignani, The Parable of Google Flu: Traps in Big Data Analysis, Science, Vol. 343, 14 March 2014, pages 1203 to 1205 - retrieved 10 August 2026, author copy at Harvard