Data study · Developer conversation on Hacker News

Hacker News Comments Saying Text Was Written by AI Rose From 16.1 to 73.5 per 100,000 Between 2023 and 2026

KGWritten by Kriszta GrenyoPublished October 20264 min read

Comments about whether a post, article or reply was written by AI grew 4.6-fold as a share of all comments since 2023. In 2026, 2.8% of popular stories drew a top-level comment raising it, up from 0.5% in 2023.

Analysis of 4,954 Hacker News comments about AI-written text, 2019 to September 2026, out of 27.3 million live comments · Hacker News public dataset on Google BigQuery, October 2026

Key findings

1
Comments about text being written by AI rose from 16.1 per 100,000 Hacker News comments in 2023 (664) to 73.5 per 100,000 in 2026 (2,141 in nine months), a 4.6-fold increase.
2
Before ChatGPT the phrase was rare: 1.8 per 100,000 comments in 2021 (66).
3
In our sample, the share of these comments that directly accused something of being AI-written rose from 38.0% in 2023 (19 of 50) to 52.9% in 2025 and 2026 (37 of 70); the rest debated detection or described their own AI use.
4
Of the 37 direct accusations sampled from 2025 and 2026, 30 (81.1%) targeted the submitted article, post or project and 2 (5.4%) targeted another commenter.
5
2.8% of stories that reached 100 points in 2026 drew a top-level comment about AI authorship (328 of 11,912), up from 0.5% in 2023 (91 of 17,767).
6
Popular AI-topic stories were three times as likely to draw one in 2026: 6.3% (139 of 2,224) versus 2.0% for other popular stories (189 of 9,688).

Summary

Picture a consultant who publishes a long blog post and shares it on a tech forum. The first reply is not about the ideas. It just says the post reads like ChatGPT wrote it. On Hacker News that reply has become much more common.

We counted comments that say or ask whether text was written by AI, from 2019 to September 2026. Per 100,000 comments, they went from 16.1 in 2023 to 73.5 in 2026. In the first nine months of 2026 alone there were 2,141.

Most direct accusations land on the thing being shared: the article, blog post or project page. The bigger risk for a business is not that readers cannot tell. It is that they assume, and stop reading.

What we measured

We used the public Hacker News table on Google BigQuery and matched live comments with phrases such as reads like, sounds like or smells like AI or ChatGPT, written or generated by AI or an LLM, AI-generated or LLM-written, and did ChatGPT write this. Rates divide matches by all live comments in each year.

Phrase matches include direct accusations and broader talk about AI-written text, such as how to detect it. Our team read random samples of 50 matched comments from 2023 and 70 from 2025 and 2026 to separate direct accusations and note their target. For popular stories, we counted stories with 100+ points that had at least one matched comment posted as a direct reply to the story.

AI-authorship talk grew 4.6-fold since 2023

73.5comments per 100,000 about text being AI-written in 2026, up from 16.1 in 2023

The first jump came with ChatGPT's launch. The second, bigger one came in 2025 and 2026 as AI-written posts became common on the web.

Comments about AI-written text per 100,000 Hacker News comments

Per 100,000 comments

Year	Per 100,000	Matched comments	All live comments
2,019	1	28	2,700,467
2,020	2.2	71	3,184,862
2,021	1.8	66	3,691,995
2,022	4.9	197	3,983,506
2,023	16.1	664	4,128,014
2,024	16.6	539	3,254,657
2,025	36.7	1,248	3,399,696
2026 (Jan to Sep)	73.5	2,141	2,911,855

Source: aisearchoptimization.net analysis of 4,954 Hacker News comments about AI-written text, 2019 to September 2026, out of 27.3 million live comments · Hacker News public dataset on Google BigQuery, October 2026

The rate barely moved between 2023 (16.1) and 2024 (16.6), then more than doubled in 2025 (36.7) and doubled again in 2026 (73.5). Matches before 2022 mostly relate to early text generators.

Accusations mostly target what gets shared

81.1%of sampled direct accusations targeted the submitted article, post or project (30 of 37)

Not every match is an accusation. Some comments debate whether em dashes or certain words give AI away, and some describe the writer's own use of AI. We read samples to see how the mix changed.

What matched comments were doing (team-coded samples)

Period	Sample size	Direct accusations	Share	Aimed at the submission	Aimed at a commenter
2023	50	19	38.0%
2025 and 2026	70	37	52.9%	30 (81.1%)	2 (5.4%)

The typical accusation says an article or project page reads like an LLM wrote it, often pointing to a flat, padded style or claims that almost make sense. Some commenters also said they stopped reading at that point.

Year	Share %	Stories with such a comment	Popular stories
2,022	0.2	26	16,908
2,023	0.5	91	17,767
2,024	0.6	78	14,068
2,025	1.4	198	14,204
2026 (Jan to Sep)	2.8	328	11,912

What this means

Readers on technical sites now check for AI writing, and when they think they see it, they say so in public. For a business, a blog post that reads like unedited AI output can cost trust before anyone gets to the point. Use AI to help draft if you like, but edit for your own voice, add real numbers and examples, and cut filler. Content that sounds like you also helps AI brand visibility, because it gives assistants something specific to quote.

Human-edited, specific content is also more likely to do well with AI assistants, which tend to favor clear sources with real facts. Our free AI visibility audit shows how your content is being picked up, and it is the starting point for our answer engine optimization services.

Limits

Phrase matching catches some comments that discuss AI-written text in general rather than accusing anything. Sample-based shares (50 and 70 comments) carry margins of error of roughly 12 to 14 points.

An accusation is not proof that the text was AI-written; we did not check.

Comment phrasing varies, so some accusations were missed.

Dead and deleted comments carry no text in the dataset, so moderated comments are not counted.

2026 covers January to September only.

Embed this research

Add the charts from this study to your page. The embed links back here.

<iframe src="https://aisearchoptimization.net/embed/data-studies/written-by-ai-comments-hacker-news" title="'Written by AI' Comments on Hacker News Up 4.6x Since 2023" width="100%" height="720" style="border:0" loading="lazy"></iframe>

Cite this study

APA

Grenyo, K. (2026). 'Written by AI' Comments on Hacker News Up 4.6x Since 2023. AI Search Optimization. https://aisearchoptimization.net/resources/data-studies/written-by-ai-comments-hacker-news

Plain link

'Written by AI' Comments on Hacker News Up 4.6x Since 2023 - AI Search Optimization - https://aisearchoptimization.net/resources/data-studies/written-by-ai-comments-hacker-news

Download the data (CSV)

Frequently asked questions

How often do Hacker News commenters say something was written by AI?

In January to September 2026, 2,141 comments did, or 73.5 per 100,000 comments. In 2023 the rate was 16.1 per 100,000.

What do these comments usually target?

In a sample of 37 direct accusations from 2025 and 2026, 30 (81.1%) targeted the submitted article, post or project and 2 (5.4%) targeted another commenter.

How many popular stories get accused of being AI-written?

In 2026, 328 of 11,912 stories with 100+ points (2.8%) drew a top-level comment about AI authorship, up from 0.5% in 2023.

Are AI-topic stories accused more often?

Yes. In 2026, 6.3% of popular AI stories drew such a comment, against 2.0% of other popular stories.

See where AI search leaves you out

Run the free AI visibility audit. If you want the fixes done for you, the plan is a flat $299/month.

Run the free audit See the $299/mo plan