Who is still reading about bitcoin

Price tells you what people paid. This tells you how many bothered to look it up — daily readers of the English Wikipedia articles for bitcoin and ether since 2015, ranked against their own eleven-year history. Crowds arrive late and leave quietly, and that shows up here first — whether it is useful lead time or just noise is the question the table below refuses to settle for you.

Bitcoin BTC

Under the 1st percentile

Forgotten — 2,266 readers a day this week, against a typical 8,174 since 7 Jul 2015. Lower than any week in the eleven years of data.

A year ago: 5,670/day
Busiest week: 250,717/day
Week ended: 13 Dec 2017

Ethereum ETH

3rd percentile

Forgotten — 629 readers a day this week, against a typical 1,651 since 7 Jul 2015.

A year ago: 1,980/day
Busiest week: 22,791/day
Week ended: 15 May 2021

Seven-day averages through 28 Jul 2026. Wikimedia publishes pageviews with a one to two day lag, so this is the freshest closed day, not today.

Attention and price, same axis

Two charts rather than two scales on one: a second y-axis would draw a relationship whose alignment we chose ourselves. Both are logarithmic — attention has run from hundreds to a quarter of a million readers a day.

Bitcoin — readers a day since 7 Jul 2015

BTC price over the same span

Ethereum — readers a day since 7 Jul 2015

ETH price over the same span

What happened next

Every day of history sorted by how much attention there was at that moment — ranked only against days before it, never against the future — then the median price change over the following three, six and twelve months. The first row is the same calculation with no filter at all: it is the number every other row has to beat to mean anything.

BTC — median change after each level of attention

Attention that day	Days	Episodes	3m	6m	12m
Every day, no filter	3,310	9	+12%	+30%	+59%
90–100 percentile the manic weeks	365	1	+57%	+138%	+87%
80–90 percentile	242	1	+45%	+41%	+19%
70–80 percentile	218	1	−4%	−11%	−23%
60–70 percentile	237	1	−4%	−18%	−37%
50–60 percentile	234	1	−8%	−26%	−7%
40–50 percentile	225	1	−7%	+9%	+59%
30–40 percentile	217	1	+6%	+35%	+61%
20–30 percentile	214	1	+23%	+34%	+79%
10–20 percentile	332	1	+28%	+48%	+111%
0–10 percentile nobody looking	1,026	3	+13%	+37%	+96%

A day counts towards a row only when it has an outcome at all three horizons, so every cell in a row describes the same set of days — the most recent year of history is not in the table yet, because its twelve months have not finished. Medians, not averages: one 2017 window would drag any average behind it. And a median hides everything interesting — across all days, the twelve-month outcomes for BTC ran from −45% to +479% between the 10th and 90th percentile. The unfiltered median was +59% over twelve months, which is the bar.

ETH — median change after each level of attention

Attention that day	Days	Episodes	3m	6m	12m
Every day, no filter	3,310	9	+12%	+24%	+37%
90–100 percentile the manic weeks	434	1	+29%	+83%	+60%
80–90 percentile	313	1	+23%	0%	−41%
70–80 percentile	269	1	+33%	+61%	−24%
60–70 percentile	380	1	−13%	−4%	+16%
50–60 percentile	391	1	−15%	−2%	+3%
40–50 percentile	359	1	+5%	+33%	+21%
30–40 percentile	332	1	+18%	+29%	+46%
20–30 percentile	354	1	+17%	+8%	+60%
10–20 percentile	335	1	+25%	+19%	+54%
0–10 percentile nobody looking	143	1	+11%	+31%	+55%

A day counts towards a row only when it has an outcome at all three horizons, so every cell in a row describes the same set of days — the most recent year of history is not in the table yet, because its twelve months have not finished. Medians, not averages: one 2017 window would drag any average behind it. And a median hides everything interesting — across all days, the twelve-month outcomes for ETH ran from −55% to +1,098% between the 10th and 90th percentile. The unfiltered median was +37% over twelve months, which is the bar.

The Episodes column is why none of this is a finding: those thousands of days overlap almost completely, and a decile holding a thousand of them is still only three or four independent year-long stretches. The buckets are not equal in size either — recent years sit low against the peaks of 2017 and 2021, so the quiet end holds far more days than the manic one.

What this means for a holder

The useful part of this page is not the table. It is the shape of the top chart: attention is not a slow-moving background level, it is a series of spikes, and the spikes land within days of the price tops — though not reliably on the same side of them: the busiest week for the bitcoin article ended three days before the December 2017 high, and the busiest week for the ether article ended four days after the May 2021 one. In December 2017 the bitcoin article was pulling 250,717 readers a day, averaged across the week. This week it is pulling 2,266.

That would be what the end of a cycle of public interest looks like from the inside, if this is one: not a crash, just quiet. Whether quiet is an opportunity or a warning is exactly the question this data cannot answer — and the table below the charts is the honest version of the attempt. For bitcoin both extremes were followed by gains and the middle was not; but the unfiltered median over twelve months is already +59%, so the quiet decile's +96% is a modest edge over simply owning it on any random day — the signature of an asset that rose over eleven years rather than of a working indicator. For ether the same shape does not appear at all: its worst twelve-month results came from the deciles just short of mania, while the middle of its range held up. Two assets, two different patterns — which is the answer.

What this page cannot tell you

Wikipedia is not the world. It measures people who wanted an encyclopaedia article, which skews towards the curious and the newly arrived, and away from anyone who already knows what bitcoin is. A holder checking a price does not appear here at all.
Eleven years is three cycles at most. The series starts in July 2015 because that is where the Wikimedia pageviews API starts. Every "what happened next" number rests on a handful of genuinely independent episodes, and the windows overlap heavily.
Attention is not demand. A news story about a hack, a court case or a national ban sends readers to the article just as reliably as a rally does. The series cannot tell those apart, and neither can we.
English only. One language edition of one encyclopaedia. Interest in markets that do not read English is invisible here.

How this page is calculated

Source. Wikimedia REST pageviews API, English Wikipedia, all-access, agent=user — automated traffic is excluded by Wikimedia before we see it. Articles: Bitcoin and Ethereum. Data available from 1 July 2015 onward.
Attention. The trailing seven-day mean of daily pageviews. Percentile is the rank of that mean among every seven-day mean in the series.
The table. For each day with at least a year of prior history, the attention rank is computed against only the days before it, then bucketed into tenths. Price changes come from the same daily closes that feed the rest of the site. Cells show the median.
Attribution. Pageview data is published by the Wikimedia Foundation under CC0.

Frequently asked questions

Is this Google Trends?

No, and the difference matters. Google's own Trends API has been in closed alpha since July 2025 and the endpoint behind the popular scraping libraries refuses automated requests, so no honest site can promise you Google search interest for free. This page uses English Wikipedia pageviews instead: the count of people who opened the "Bitcoin" or "Ethereum" article on a given day, published openly by the Wikimedia Foundation with automated traffic already filtered out. It is a different measure of the same thing — how many people are curious enough to go read.

Why the seven-day average?

Because the daily series has a weekday rhythm: fewer people read encyclopaedias at the weekend. A seven-day average removes that without smoothing away the spikes that actually matter, and every number on this page — today's reading, the percentile, the table — is computed on the same seven-day basis.

What does the percentile mean?

Where the last seven days rank among every seven-day stretch since 7 Jul 2015. A reading in the 5th percentile means attention has been higher than this on 95% of days in eleven years. It is a rank inside this one series, not a share of the world's population.

Does low attention mean it is time to buy?

No, and the table below is the reason rather than the proof. For bitcoin it shows the quiet end and the manic end both preceding gains while the middle did not — a shape that says more about an asset which rose over eleven years than about attention. For ether the same table does not hold that shape at all: its worst twelve-month outcomes came from the deciles just short of mania — the 70th to 90th percentile of attention — while the middle of its range was mildly positive. Two assets, two different patterns, one price history each. That is what a coincidence looks like, and it is why nothing here is a signal. See what this site refuses to do.

How fresh is it?

Wikimedia publishes with a one to two day lag, so the last point is normally yesterday or the day before. Our cron collects the series daily and the page states its own as-of date above.

Data through 28 Jul 2026 · All tools