Picture a firehose pointed at a teacup.
The firehose is your email. It gets stronger every year. The teacup is your attention, and it has been the same size for as long as anyone has measured it. That picture is the whole problem, and the rest of this piece is the data behind it.
Two numbers carry the argument. Worldwide email volume grew from 247 billion messages a day in 2009 to 376 billion in 2025, and it keeps compounding at about 4% a year (Radicati Group via Statista). The average adult reads at 238 words a minute, a number that has not moved (Brysbaert meta-analysis, Journal of Memory and Language, 2019). One line goes up and to the right. The other is a flat line. They never reconcile.
The short version
- Email volume compounds at ~4% a year, rising toward 8%+ as AI removes the cost of sending.
- Human attention is flat: ~238 words a minute, ~28% of the workweek on email, unchanged for over a decade.
- ~45% of all email is spam, and ~76% of the legitimate rest is noise. Roughly 1 in 10 messages is a real person writing to you.
- By 2036, volume could be 2.2 to 3 times today's while attention stays at 1x. The gap is the entire problem.
1. The firehose: volume is compounding
The standard industry series for email volume comes from the Radicati Group, distributed through Statista. It is proprietary and not independently audited, so treat the absolute numbers as directional. But the shape is consistent across every recent year, and the shape is what matters.
| Year | Emails / Day | Email Users |
|---|---|---|
| 2017 | 269.0 billion | 3.72 billion |
| 2019 | 293.6 billion | 3.93 billion |
| 2021 | 319.6 billion | 4.15 billion |
| 2023 | 347.3 billion | 4.37 billion |
| 2025 | 376.4 billion | 4.59 billion |
| 2026 (forecast) | 392.5 billion | 4.73 billion |
| 2028 (forecast) | 424.2 billion | 4.97 billion |
Volume grows about 4.1% a year. Users grow about 3%. So even before you account for anything else, email per person is climbing. Annualized, 2025's 376 billion a day works out to roughly 137 trillion emails a year. For the wider set of benchmarks behind these figures, see our 50+ email statistics reference and the full 2026 State of the Inbox report.
2. The teacup: attention is flat
Now the other line. The one that does not move.
Reading speed is the hard ceiling. Marc Brysbaert's 2019 meta-analysis pooled 190 studies and 18,573 participants and put adult silent reading at 238 words a minute for non-fiction, with comprehension falling apart above roughly 500. You cannot train your way out of this at scale. It is a property of human cognition, not a habit you can optimize.
Time spent on email tells the same story. Adobe's Email Usage Study tracked daily email time falling from 465 minutes in 2016 to 321 in 2021. McKinsey's benchmark, that the average interaction worker spends 28% of the workweek managing email, dates to 2012 and has not meaningfully risen since. There are only so many minutes in a day, and they were already full a decade ago.
A business user receives 120 or more emails a day (Radicati: 126 sent and received; cloudHQ: ~121 received). At a realistic minute or two each to triage, read, and decide, that is two to four hours a day before you have done any actual work. The teacup is full. The firehose does not care.
3. Most of what arrives is not a person
Here is the part that should bother you most. The flood is mostly machines.
About 45% of all email is spam (Kaspersky, 2025). Spam peaked near 88% around 2009 and has fallen as a share, but because total volume rose, the absolute amount of spam stayed roughly flat at 160 billion or more messages a day. Then there is the legitimate mail. cloudHQ's 2025 analysis found only about 24% of received messages are important, and the other 76% are newsletters, marketing, transactional receipts, automated notifications, and unnecessary CCs.
| Segment of all email | Approx. share | Basis |
|---|---|---|
| Spam / junk | ~45% | Kaspersky, 2025 |
| Bulk legit (marketing, transactional, notifications, newsletters) | ~45% | cloudHQ "76% noise" |
| Genuine human one-to-one | ~10% | Analyst synthesis (labelled estimate) |
Stack it up and genuine one-to-one human email is roughly 1 in 10 messages. That last figure is a synthesis, not a single measured stat, so I am labelling it as an estimate. But the direction is not in doubt. The reason is simple economics: a person can only write so many emails a day, so human mail is capped by human labor. Machine mail has no such cap. This is the same point I make through Claude Shannon's lens in the essay on decoding the inbox. The bits arrive perfectly. The meaning is buried.
4. AI bends the curve sharply upward
For 15 years, email volume grew on a steady ~4% line because sending still took human effort. That cap is coming off.
AI sales agents now research and personalize cold outreach at a scale no team could match. One SDR can research ten prospects an hour. An AI agent can research ten thousand in minutes. On the hostile side, Hoxhunt's 2026 Phishing Trends Report found AI-generated phishing soared from 4% to 56% of attacks over the December 2025 holiday season, a 14x surge in filter-bypassing mail. cloudHQ now models automated email growing more than 9% a year while human-written mail crawls.
| Scenario for 2036 | Emails / Day | Assumption |
|---|---|---|
| Baseline | ~581 billion | Historical ~4% CAGR continues |
| AI-accelerated | ~810–890 billion | Automated mail grows ~8%/yr, human ~2%/yr |
Indexed to 2017, that puts email volume at 2.2 times its starting point by 2036 in the baseline, and around 3 times under the AI scenario. Users reach about 1.7x. Attention stays at 1x. The firehose gets a bigger pump. The teacup stays a teacup.
5. The gap is the product
Lay the two lines on one chart and the story is a single shape: volume rockets, users plod, attention flatlines. The widening wedge between the volume line and the attention line is not an abstraction. It is the hours you lose every week reading mail that a machine sent to another machine's specifications, hoping the one message that matters is somewhere in the pile.
You cannot read faster. You cannot add hours to the day. The volume is not going to politely slow down. So there is only one variable left to change: stop reading the channel, and read what it meant instead.
That is the bet behind MailOver. An AI reads every message, throws out the 88% that is noise, and hands you the rest as a short briefing: the action items, the deadlines, the decisions waiting on you. Newsletters and receipts get categorized and kept out of your way automatically. Instead of 120 messages, you get one daily briefing. Presidents and CEOs solved this problem decades ago. They hired a secretary to read the channel for them. The only new thing is that the secretary now costs a few dollars a month instead of a salary.
Key takeaways
- Email volume grew from 247B/day (2009) to 376B/day (2025) and compounds at ~4% a year.
- Adult reading speed is ~238 wpm and time-per-person on email is flat to declining.
- ~45% of email is spam; ~76% of the legitimate rest is noise; ~10% is genuine human one-to-one.
- AI is pushing volume toward 800B+ a day by 2036 while attention stays fixed.
- The only movable variable is the interface: delegate the reading and consume a briefing.
Methodology and sources
Figures are compiled from the primary sources below. Where a value is an estimate or a synthesis rather than a single measured datapoint, it is labelled as such. Radicati's series is widely cited but proprietary, and its older forecasts overshot before being revised, so absolute numbers are directional.
- Radicati Group (via Statista), 2017–2028: worldwide email volume and user counts.
- Marc Brysbaert, Journal of Memory and Language, vol. 109 (2019): 238 wpm average adult silent reading speed, from 190 studies and 18,573 participants.
- McKinsey Global Institute (2012): 28% of the workweek spent managing email.
- Adobe Email Usage Study: time-per-person on email (US, self-reported), 2016–2021.
- Kaspersky Securelist, 2025: spam at ~45% of all email.
- cloudHQ, 2025–2026: ~24% of mail important / ~76% noise; automated-email growth above 9%/yr; 2030 volume estimate.
- Hoxhunt 2026 Phishing Trends Report: AI phishing surge from 4% to 56% over the December 2025 holiday season.
- IBM Cost of a Data Breach Report 2025 (Ponemon): 1 in 6 breaches involved attackers using AI.
Projections to 2036 are clearly labelled as scenarios. The baseline applies the ~4% historical CAGR; the AI-accelerated scenario assumes the automated component grows ~8%/yr and human mail ~2%/yr, which lands the 2030 figure close to cloudHQ's independent estimate.
Frequently asked questions
How fast is email volume growing?
Worldwide email volume has grown about 4.1% a year for the last decade, from 269 billion messages a day in 2017 to 376 billion in 2025, with Radicati forecasting 392.5 billion in 2026. AI is set to break that curve. cloudHQ models automated email growing more than 9% a year, which points to 800 to 890 billion messages a day by 2036 under an AI-accelerated scenario, against a baseline of about 581 billion at the historical rate.
Is human attention keeping up with email volume?
No. Adult silent reading speed averages about 238 words a minute and has not changed. Time spent on email per person is flat to declining, and the McKinsey benchmark of 28% of the workweek on email has held since 2012. Email users grow only about 3% a year. Volume per person rises every year while the minutes available to read it do not move.
How much of the inbox is actually written by a human?
A minority of it. About 45% of all email is spam, and of the legitimate remainder roughly 76% is noise: newsletters, marketing, transactional mail, automated notifications, and unnecessary CCs. Genuine one-to-one human email is roughly 1 in 10 messages, because human email is capped by human labor while machine email is not.
Why does AI make the email problem worse before it makes it better?
AI removes the labor cap on sending. AI sales agents can personalize outreach to thousands of prospects in minutes, and AI-generated phishing surged from 4% to 56% of attacks over the December 2025 holiday season. The same generative AI that floods the inbox is the only thing that can decode it on the receiving side. AI on the sender's side is the threat. AI on your side is the cure.
