The Rise of the Em—Dash

Two signals, one story

The blue bars count any issue with an em-dash anywhere in the title or body — that's the flood. The green line counts the much narrower set of issues that talk about em-dashes by name (“em dash” or “em-dash”) — bug reports, feature requests, and complaints filed because of em-dashes. From a 2021 baseline of ~25/month, the about-em-dash signal roughly doubles by 2025 and triples in late 2025, peaking at 141 in 2025-09. Em-dashes are no longer just appearing — they're causing enough friction to be noticed.

What happened?

From 2021 through early 2024 the baseline hovers around 3,000–8,000 issues per month, with occasional spikes from automated bots and template-text dumps. In mid-2025 the curve breaks loose: July hits six figures, August clears 430,000, and by April 2026 a single month logs over half a million issues containing em-dashes — roughly a hundredfold rise from the pre-LLM baseline.

The query

Run against the public githubarchive BigQuery dataset, counting IssuesEvent actions of type opened whose title or body contains an em-dash character or the literal phrase “em dash”.

SELECT
  FORMAT_DATE('%Y-%m', DATE(created_at)) AS month,
  COUNT(*) AS issue_count
FROM
  `githubarchive.month.*`
WHERE
  type = 'IssuesEvent'
  AND JSON_EXTRACT_SCALAR(payload, '$.action') = 'opened'
  AND (
    LOWER(JSON_EXTRACT_SCALAR(payload, '$.issue.title')) LIKE '%em dash%'
    OR LOWER(JSON_EXTRACT_SCALAR(payload, '$.issue.body')) LIKE '%em dash%'
    OR JSON_EXTRACT_SCALAR(payload, '$.issue.body') LIKE '%—%'
  )
  AND _TABLE_SUFFIX BETWEEN '2020_01' AND '2026_04'
GROUP BY month
ORDER BY month

The tighter query (green line)

Drops the em-dash-character match entirely — only counts issues whose text literally contains the phrase “em dash” or “em-dash”. These are issues filed because of em-dashes, not just ones that happen to contain one.

SELECT
  FORMAT_DATE('%Y-%m', DATE(created_at)) AS month,
  COUNT(*) AS issue_count
FROM
  `githubarchive.month.*`
WHERE
  type = 'IssuesEvent'
  AND JSON_EXTRACT_SCALAR(payload, '$.action') = 'opened'
  AND (
    LOWER(JSON_EXTRACT_SCALAR(payload, '$.issue.title')) LIKE '%em-dash%'
    OR LOWER(JSON_EXTRACT_SCALAR(payload, '$.issue.title')) LIKE '%em dash%'
    OR LOWER(JSON_EXTRACT_SCALAR(payload, '$.issue.body')) LIKE '%em-dash%'
    OR LOWER(JSON_EXTRACT_SCALAR(payload, '$.issue.body')) LIKE '%em dash%'
  )
  AND _TABLE_SUFFIX BETWEEN '2020_01' AND '2026_04'
GROUP BY month
ORDER BY month