Data study · Developer conversation on Hacker News
Only 2.1% of the Links Shared in Hacker News AI Search Debates Point to Research Papers
Of 5,740 links in 3,370 AI search stories and their comment threads, 119 went to research papers. News outlets supplied 20.1% and independent sites 35.4%. GitHub was the single most-linked site.
Analysis of 5,740 links from 3,370 Hacker News AI search stories and their comment threads, January 2023 to September 2026 · Hacker News public dataset on Google BigQuery, October 2026
Key findings
- 1
Research papers made up 2.1% of the 5,740 links shared in Hacker News AI search debates (119), and their share fell from 3.4% in 2023 (61 of 1,813) to 1.2% in 2026 (11 of 953).
- 2
News and trade media supplied 20.1% of links (1,155), mostly as the story being discussed: 947 of their 1,155 links were story submissions.
- 3
Independent blogs, small companies and other sites made up the largest group at 35.4% (2,034), rising from 27.1% of links in 2023 to 42.1% in 2026.
- 4
Links to the AI and search companies themselves were 17.6% of the total (1,011), and two thirds of those (677) were shared by commenters rather than submitted as stories.
- 5
GitHub was the most-linked domain with 440 links, followed by Twitter (162), Hacker News itself (160) and The Verge (150).
- 6
Code and model hosting sites doubled their share, from 6.9% of links in 2023 (125) to 15.5% in 2026 (148).
Summary
Say a founder wants to understand whether AI search will hurt their traffic, so they read the Hacker News threads where developers argue about it. What sources are those arguments built on? Mostly news write-ups, company announcements and people's own blog posts. Very little peer-reviewed or preprint research.
We pulled every link from 3,370 AI search stories posted since January 2023, plus the links people shared inside the comment threads. Of 5,740 links, only 119 went to research papers.
The mix has also shifted. Independent blogs and small company sites now provide the biggest share, while news outlets and research both lost ground. The links suggest opinion in this space leans on first-hand write-ups and tool repos more than on studies.
What we measured
We started from the 3,370 AI search stories in our related engagement study. For each, we took the story's own URL and every link inside the comments on that thread that are in our working extract (comments mentioning search or AI terms, which covers most of the discussion). That gave 3,130 story links and 2,610 comment links.
Each link was reduced to its domain and grouped into source types: research papers (arXiv, ACM, journals and university sites), news and trade media, AI and search company sites, social and video platforms, forums and archives, code and model hosting, Wikipedia, platform-hosted blogs, and all other sites. The last group is mostly independent blogs and small company sites. We name only widely known platforms.
Research barely shows up
For a technical crowd, the AI search debate leans surprisingly little on studies. Most links point to news coverage, company pages and people's own write-ups.
% of links
| Source type | Share % | Links | Submitted as story | Shared in comments |
|---|---|---|---|---|
| Other sites (mostly independent blogs and small companies) | 35.4 | 2,034 | 1,237 | 797 |
| News and trade media | 20.1 | 1,155 | 947 | 208 |
| AI and search company sites | 17.6 | 1,011 | 334 | 677 |
| Social, video, forums and archives | 11.8 | 679 | 208 | 471 |
| Code and model hosting | 8.8 | 505 | 270 | 235 |
| Wikipedia | 2.1 | 120 | 2 | 118 |
| Research papers | 2.1 | 119 | 41 | 78 |
| Blogs on publishing platforms | 2 | 117 | 91 | 26 |
Source: aisearchoptimization.net analysis of 5,740 links from 3,370 Hacker News AI search stories and their comment threads, January 2023 to September 2026 · Hacker News public dataset on Google BigQuery, October 2026
News outlets mostly appear as the story under discussion. Company sites, Wikipedia and research papers mostly appear in comments, where people back up a point. Even there, research papers made up 78 of 2,610 comment links (3.0%).
Independent voices are taking a bigger share
The mix of sources changed each year as AI search moved from a news event to an everyday tool.
| Source type | 2023 | 2024 | 2025 | 2026 (Jan to Sep) |
|---|---|---|---|---|
| Other sites (mostly independent blogs and small companies) | 27.1% | 36.4% | 40.2% | 42.1% |
| News and trade media | 22.8% | 21.0% | 18.7% | 16.1% |
| AI and search company sites | 19.3% | 17.6% | 17.8% | 14.2% |
| Social, video, forums and archives | 15.4% | 11.9% | 10.1% | 7.8% |
| Code and model hosting | 6.9% | 7.9% | 7.7% | 15.5% |
| Wikipedia | 2.5% | 2.5% | 2.1% | 0.8% |
| Research papers | 3.4% | 1.7% | 1.5% | 1.2% |
| Blogs on publishing platforms | 2.6% | 1.1% | 2.0% | 2.4% |
| All links (count) | 1,813 | 1,426 | 1,548 | 953 |
Source: aisearchoptimization.net analysis of 5,740 links from 3,370 Hacker News AI search stories and their comment threads, January 2023 to September 2026 · Hacker News public dataset on Google BigQuery, October 2026
News media fell from 22.8% of links in 2023 to 16.1% in 2026, and research from 3.4% to 1.2%. Independent sites rose from 27.1% to 42.1%, and code and model hosting from 6.9% to 15.5%, as more threads centered on tools people built.
The 25 most-linked domains
These are the individual domains linked most often across stories and comments. A site can appear because it was praised, criticized or simply cited.
| Rank | Domain | Links |
|---|---|---|
| 1 | github.com | 440 |
| 2 | twitter.com | 162 |
| 3 | news.ycombinator.com | 160 |
| 4 | theverge.com | 150 |
| 5 | phind.com | 118 |
| 6 | techcrunch.com | 113 |
| 7 | wikipedia.org | 108 |
| 8 | reddit.com | 101 |
| 9 | arstechnica.com | 101 |
| 10 | youtube.com | 100 |
| 11 | perplexity.ai | 95 |
| 12 | arxiv.org | 78 |
| 13 | blog.google | 72 |
| 14 | bloomberg.com | 65 |
| 15 | reuters.com | 54 |
| 16 | medium.com | 48 |
| 17 | kagi.com | 47 |
| 18 | wsj.com | 46 |
| 19 | wired.com | 46 |
| 20 | google.com | 45 |
| 21 | theregister.com | 45 |
| 22 | chatgpt.com | 34 |
| 23 | chromewebstore.google.com | 33 |
| 24 | searchengineland.com | 31 |
| 25 | cnbc.com | 30 |
Source: aisearchoptimization.net analysis of 5,740 links from 3,370 Hacker News AI search stories and their comment threads, January 2023 to September 2026 · Hacker News public dataset on Google BigQuery, October 2026
Two kinds of sites dominate the top 25: tech news outlets and the AI and search products being discussed. Hacker News links to itself 160 times, as people point to earlier threads. X links under its newer domain added 22 more to Twitter's 162.
What this means
The sources that shape opinion about AI search among technical buyers are news write-ups, product announcements and hands-on posts from people who tried things. If you publish something useful about AI search, a clear first-hand write-up with real data is the format this audience shares.
Something similar is likely true of how AI tools pick sources: specific, firsthand content with real numbers tends to get cited. The pages AI answers skip are where a GEO agency should put original data first, and a free AI visibility audit shows you which ones they are.
Limits
Comment links come only from comments in our working extract, which keeps comments that mention search, AI or related terms. Comments in these threads that mention none of those terms were not scanned, so comment links are undercounted.
AI search stories were matched on title keywords; a 60-title check found about 87% clearly on topic.
Domain grouping is rule-based. Some news sites or research hosts may have landed in the other group.
A link shows a source was referenced, not that it was trusted or agreed with.
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/websites-linked-hacker-news-ai-search-debates" title="Only 2.1% of Links in HN AI Search Debates Cite Research" width="100%" height="720" style="border:0" loading="lazy"></iframe>
Cite this study
APA
Grenyo, K. (2026). Only 2.1% of Links in HN AI Search Debates Cite Research. AI Search Optimization. https://aisearchoptimization.net/resources/data-studies/websites-linked-hacker-news-ai-search-debates
Plain link
Only 2.1% of Links in HN AI Search Debates Cite Research - AI Search Optimization - https://aisearchoptimization.net/resources/data-studies/websites-linked-hacker-news-ai-search-debates
Frequently asked questions
What sources do developers cite when debating AI search?
Of 5,740 links in Hacker News AI search threads, 35.4% went to independent blogs, small companies and other sites, 20.1% to news and trade media and 17.6% to AI and search company sites. Research papers were 2.1%.
Which website is linked most in Hacker News AI search threads?
GitHub, with 440 links, followed by Twitter (162), Hacker News itself (160) and The Verge (150).
Are research papers becoming more common in these debates?
No. Their share fell from 3.4% of links in 2023 to 1.2% in 2026.
Where do news links appear?
Mostly as the story itself. 947 of 1,155 news links were story submissions; only 208 were shared inside comments.
Related studies
- Developer conversation on Hacker News'Written by AI' Comments on Hacker News Up 4.6x Since 2023
- Developer conversation on Hacker Newsllms.txt Splits Developers: 54% For, 46% Against
- Developer conversation on Hacker News'Slop' Now Shows Up in 1 of Every 149 Hacker News Comments
- Developer conversation on Hacker NewsAI SEO Draws 3x the Positive Developer Comments of SEO
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.