T
Every email advice column in existence opens with the salutation. It is treated as the load-bearing decision — get it wrong and you are too stiff, too familiar, too American, too much. Entire consultancies exist to tell you whether to write Dear Ms Patel or Hi Priya.
We counted the openings of 71,216 real emails. The overwhelming majority of them skip the question entirely.
This is the sister study to our analysis of how those same emails end. Same 71,216 messages, opposite end of the page: who says hi, who says hello, who says Dear — and why the most common greeting in real correspondence is no greeting at all.
Across the two human corpora in our dataset — the Git mailing list (44,627 messages, 2022–2026) and Enron's internal mail (12,091 messages, 1999–2002) — 75.1% of messages open with no greeting of any kind. Another 15.6% open with a bare name: Priya, and then straight into it.
That leaves 9.3% of email that begins the way the advice columns assume all email begins.
This is not new, and it is not a sign of decline. Joan Waldvogel's 2007 study of 515 workplace emails from two New Zealand organisations found a 59% no-greeting rate at one of them — a government-funded education provider — against just 17% at a manufacturing plant a few miles away.1 Two organisations, same country, same year, same technology, a three-and-a-half-fold difference. Her conclusion was that workplace culture predicted greeting behaviour better than status, social distance or gender did.
Our corpora measured directly; Waldvogel (2007), Rygg (2021) and Al Rousan et al. (2024) as published. Definitions vary slightly between studies — read the spread, not the decimal.
The spread runs from 3% to 81%. Arabic academic email, as measured by Al Rousan and colleagues across 200 messages to a professor's inbox, is almost universally greeted: 97% carried a salutation, and only 3% had none.2 Norwegian workplace email sits at 13.9% no-opening across 927 messages.3 The Git mailing list sits at 80.8%.
There is no such thing as "how people open emails." There is only how a particular group opens emails to each other.
Narrow to the messages that actually use a greeting word, and the picture sharpens fast
On the Git list, Hi takes 67.9% of all greetings — it is not the most popular option so much as the default one, with Hello a distant second at 20.4%. Dear is 5.5% and Hey, the supposed marker of millennial informality, is 2.3%.
Enron is the more interesting case, because it is the closest thing we have to ordinary corporate email. There, greetings split almost evenly between Hi at 35.2% and Dear at 28.8%. A quarter-century ago, in an American office, Dear was still a live option in internal mail. On a 2026 open-source list it has effectively disappeared.
The self-reported surveys agree on the ranking, and disagree wildly on the magnitude. Preply's December 2022 survey of 1,005 Americans found 67% claiming to use Hi, 54% Good morning/afternoon/evening, 48% Hello, 19% Dear and 15% Hey.4 Perkbox's 1,928 UK workers ranked Hi best at 49% and Good morning/afternoon at 48%.5
Note what that gap implies. Over half of surveyed people say they use Good morning. In 56,718 human emails we found it in 0.2% of messages. People are describing an email they have written a handful of times, not the ones they send every day.
There is one prior large-scale measurement worth putting beside ours. In December 2017 Boomerang analysed the opening lines of 300,000+ threads and cross-referenced them against reply rates. Every opening beat the 47.5% baseline, and the informal ones won: Hey 64.0%, Hello 63.6%, Hi 62.7%, against Greetings at 57.2% and Dear at 56.5%.7 Boomerang were careful to note the limitation, and it is the same one we carry: their corpus was public mailing list archives, which is close to the least formal email population in existence.
As with sign-offs, the fraud corpora gave us our clearest signal — and it runs exactly opposite to the human pattern.
Real correspondents say Hi. Spam and phishing say Hello and Dear. Among all analysed messages, Hello appears 3.3 times more often in fraudulent mail than in human mail, and Dear 3.3 times more often. Hi runs the other way: humans use it nearly twice as often as scammers do.
In our phishing corpus, Dear is the single most common greeting — 33.2% of the messages that use one. Among Git list contributors it is 5.5%.
Human n = 56,718 messages (Git list + Enron). Fraud n = 14,484 messages (Enron-Spam + Phishing Pot). Ratio is fraud share ÷ human share.
The reason is structural rather than psychological. Dear and Hello survive in bulk mail because they work without a name, or with a name mail-merged in. Hi is what you write when you already know who you are writing to. A greeting that scales is, definitionally, a greeting that doesn't know you.
Which leaves the advice industry in an awkward position. Perkbox's respondents rated no greeting the single worst way to open an email, at 53%.5 It is also, by a distance, the most common thing real colleagues actually do.
The received narrative is that email formality is in freefall — Gen Z killed the salutation, Slack finished the job. Barclays' 2023 survey of 2,000+ UK adults found 73% saying workplace communication has become more casual, and 71% crediting Gen Z for it.6
Our one longitudinal window does not show that. Tracking the Git mailing list year by year from 2022 to September 2026 — the same community, the same norms, roughly 9,000 messages a year — greeting use is rising.
Five years of openings
2026 covers 1 January to 22 September. Shares are of all analysed messages that year, not of greeters.
Hi grew by 63% in relative terms across five years. The no-greeting share fell from 82.5% to 80.2%. These are small absolute movements in a community with strong conventions, but they point the opposite way from the decline story — and they are measured rather than recalled.
A plausible reading: as a mailing list absorbs more newcomers and more first-time contributors, the share of messages that are approaches rather than continuations goes up, and approaches need greetings. Formality here is not a generational trait. It is a function of how well you already know the person.
Messages were parsed from raw RFC822 text. We stripped quoted lines, forwarded and original-message blocks, patch payloads and signature files, leaving only composed text. The first two non-empty lines were matched against salutation patterns anchored as prefixes, so Hi Bob, can you look at this? is caught as readily as Hi Bob, on its own line. A bare name plus comma is counted separately from a greeting word. Messages with fewer than fifteen words of prose were excluded, as were exact duplicates of substantial messages.
Two corpora carry most of the weight, and neither represents email generally. The Git list is a technical, English-language, heavily male open-source community with unusually strong reply conventions; Enron is one American energy company, and the public release of it is lowercased and whitespace-tokenised. Client-facing, junior-to-senior and cross-cultural email — exactly where greetings survive — is under-represented. Treat our no-greeting figures as an upper bound for informal contexts, and the published cross-cultural studies in Figure 1 as the corrective.
The phishing corpus is the smallest at 1,708 classified messages and skews toward campaigns rather than individual senders, so its greeting shares carry the widest error. We also evaluated and rejected the released Clinton State Department emails (OCR fragments interleaved with FOIA stamps) and a 1990s mailing list archive that leaked mail headers into message bodies.
We counted the openings of 71,216 real emails. The overwhelming majority of them skip the question entirely.
This is the sister study to our analysis of how those same emails end. Same 71,216 messages, opposite end of the page: who says hi, who says hello, who says Dear — and why the most common greeting in real correspondence is no greeting at all.
The 75% who say nothing
Across the two human corpora in our dataset — the Git mailing list (44,627 messages, 2022–2026) and Enron's internal mail (12,091 messages, 1999–2002) — 75.1% of messages open with no greeting of any kind. Another 15.6% open with a bare name: Priya, and then straight into it.
That leaves 9.3% of email that begins the way the advice columns assume all email begins.
This is not new, and it is not a sign of decline. Joan Waldvogel's 2007 study of 515 workplace emails from two New Zealand organisations found a 59% no-greeting rate at one of them — a government-funded education provider — against just 17% at a manufacturing plant a few miles away.1 Two organisations, same country, same year, same technology, a three-and-a-half-fold difference. Her conclusion was that workplace culture predicted greeting behaviour better than status, social distance or gender did.
Our corpora measured directly; Waldvogel (2007), Rygg (2021) and Al Rousan et al. (2024) as published. Definitions vary slightly between studies — read the spread, not the decimal.
The spread runs from 3% to 81%. Arabic academic email, as measured by Al Rousan and colleagues across 200 messages to a professor's inbox, is almost universally greeted: 97% carried a salutation, and only 3% had none.2 Norwegian workplace email sits at 13.9% no-opening across 927 messages.3 The Git mailing list sits at 80.8%.
There is no such thing as "how people open emails." There is only how a particular group opens emails to each other.
When people do greet, they say Hi
Narrow to the messages that actually use a greeting word, and the picture sharpens fast
On the Git list, Hi takes 67.9% of all greetings — it is not the most popular option so much as the default one, with Hello a distant second at 20.4%. Dear is 5.5% and Hey, the supposed marker of millennial informality, is 2.3%.
Enron is the more interesting case, because it is the closest thing we have to ordinary corporate email. There, greetings split almost evenly between Hi at 35.2% and Dear at 28.8%. A quarter-century ago, in an American office, Dear was still a live option in internal mail. On a 2026 open-source list it has effectively disappeared.
The self-reported surveys agree on the ranking, and disagree wildly on the magnitude. Preply's December 2022 survey of 1,005 Americans found 67% claiming to use Hi, 54% Good morning/afternoon/evening, 48% Hello, 19% Dear and 15% Hey.4 Perkbox's 1,928 UK workers ranked Hi best at 49% and Good morning/afternoon at 48%.5
Note what that gap implies. Over half of surveyed people say they use Good morning. In 56,718 human emails we found it in 0.2% of messages. People are describing an email they have written a handful of times, not the ones they send every day.
There is one prior large-scale measurement worth putting beside ours. In December 2017 Boomerang analysed the opening lines of 300,000+ threads and cross-referenced them against reply rates. Every opening beat the 47.5% baseline, and the informal ones won: Hey 64.0%, Hello 63.6%, Hi 62.7%, against Greetings at 57.2% and Dear at 56.5%.7 Boomerang were careful to note the limitation, and it is the same one we carry: their corpus was public mailing list archives, which is close to the least formal email population in existence.
Hello and Dear are phishing tells
As with sign-offs, the fraud corpora gave us our clearest signal — and it runs exactly opposite to the human pattern.
Real correspondents say Hi. Spam and phishing say Hello and Dear. Among all analysed messages, Hello appears 3.3 times more often in fraudulent mail than in human mail, and Dear 3.3 times more often. Hi runs the other way: humans use it nearly twice as often as scammers do.
In our phishing corpus, Dear is the single most common greeting — 33.2% of the messages that use one. Among Git list contributors it is 5.5%.
Human n = 56,718 messages (Git list + Enron). Fraud n = 14,484 messages (Enron-Spam + Phishing Pot). Ratio is fraud share ÷ human share.
The reason is structural rather than psychological. Dear and Hello survive in bulk mail because they work without a name, or with a name mail-merged in. Hi is what you write when you already know who you are writing to. A greeting that scales is, definitionally, a greeting that doesn't know you.
Which leaves the advice industry in an awkward position. Perkbox's respondents rated no greeting the single worst way to open an email, at 53%.5 It is also, by a distance, the most common thing real colleagues actually do.
The list is getting friendlier, not colder
The received narrative is that email formality is in freefall — Gen Z killed the salutation, Slack finished the job. Barclays' 2023 survey of 2,000+ UK adults found 73% saying workplace communication has become more casual, and 71% crediting Gen Z for it.6
Our one longitudinal window does not show that. Tracking the Git mailing list year by year from 2022 to September 2026 — the same community, the same norms, roughly 9,000 messages a year — greeting use is rising.
Five years of openings
| Year | Messages | No greeting | Hi | Name only | Hello |
|---|---|---|---|---|---|
| 2022 | 9,255 | 82.5% | 4.9% | 10.2% | 1.4% |
| 2023 | 7,167 | 81.2% | 5.7% | 10.0% | 2.0% |
| 2024 | 9,808 | 80.2% | 6.5% | 9.8% | 2.5% |
| 2025 | 9,717 | 80.1% | 6.8% | 9.8% | 2.0% |
| 2026 | 8,678 | 80.2% | 8.0% | 9.2% | 1.6% |
2026 covers 1 January to 22 September. Shares are of all analysed messages that year, not of greeters.
Hi grew by 63% in relative terms across five years. The no-greeting share fell from 82.5% to 80.2%. These are small absolute movements in a community with strong conventions, but they point the opposite way from the decline story — and they are measured rather than recalled.
A plausible reading: as a mailing list absorbs more newcomers and more first-time contributors, the share of messages that are approaches rather than continuations goes up, and approaches need greetings. Formality here is not a generational trait. It is a function of how well you already know the person.
Method and limitations
Method
Messages were parsed from raw RFC822 text. We stripped quoted lines, forwarded and original-message blocks, patch payloads and signature files, leaving only composed text. The first two non-empty lines were matched against salutation patterns anchored as prefixes, so Hi Bob, can you look at this? is caught as readily as Hi Bob, on its own line. A bare name plus comma is counted separately from a greeting word. Messages with fewer than fifteen words of prose were excluded, as were exact duplicates of substantial messages.
Limitations
Two corpora carry most of the weight, and neither represents email generally. The Git list is a technical, English-language, heavily male open-source community with unusually strong reply conventions; Enron is one American energy company, and the public release of it is lowercased and whitespace-tokenised. Client-facing, junior-to-senior and cross-cultural email — exactly where greetings survive — is under-represented. Treat our no-greeting figures as an upper bound for informal contexts, and the published cross-cultural studies in Figure 1 as the corrective.
The phishing corpus is the smallest at 1,708 classified messages and skews toward campaigns rather than individual senders, so its greeting shares carry the widest error. We also evaluated and rejected the released Clinton State Department emails (OCR fragments interleaved with FOIA stamps) and a 1990s mailing list archive that leaked mail headers into message bodies.
Sources
- Joan Waldvogel, "Greetings and Closings in Workplace Email", Journal of Computer-Mediated Communication 12(2), 2007. 515 emails, two New Zealand organisations.
- Al Rousan et al., "Good Morning, My Dear Dr.", World Journal of English Language, 2024. 200 Arabic academic emails.
- Kristin Rygg, "Openings and Closings in Workplace Emails", Hermes 61, 2021. 927 Norwegian workplace emails, 2015–2019.
- Preply, "Email Greetings and Sign-Offs", 1,005 US adults, fielded 9 December 2022.
- Perkbox Insights, "The most annoying things in an email, ranked", 1,928 UK respondents, January 2020.
- Barclays / Censuswide, "Say goodbye to Yours Truly", 2,000+ UK adults, fieldwork August 2023.
- Boomerang, "How to Start an Email", December 2017. 300,000+ threads; reply rates by opening word.
- Corpora: Git mailing list mirror; Enron-Spam; Phishing Pot.