Short answer: In developer-media buyer prompts, the outlets AI engines recommend are not the domains they link to when sources appear. Among 191 UI observations that included at least one cited URL, Hashnode is named on 65.4% of answers but hashnode.com appears in the source list on 5.2%. Dev.to is named on 69.6% but dev.to is linked on 26.7% — 88 cited answers mention Dev.to without linking to it. In Plain English is the exception: 44.0% mention rate and 41.4% citation rate on the same subset, with resources.plainenglish.io leading the category citation leaderboard at 9.5% of all linked URLs while the publication brand holds 3.1% of mention share corpus-wide.
Measuring “AI visibility” from citations alone will over-weight In Plain English and under-weight Hashnode relative to what models actually recommend.
What we analysed
We used Obsurfable Explorer’s research corpus: buyer-style prompts tagged Technology · Developer Media — where to publish technical writing, which developer blogs to follow, and which communities engineers use.
For each prompt we keep the latest observation per platform family (ChatGPT, Gemini, Claude, Perplexity, Grok), preferring consumer chat captures over API batch runs when both exist. We compared:
- Mention rates — brand appears anywhere in the ordered recommendation list extracted from the answer.
- Citation rates — cited URL hostname matches the outlet (e.g.
dev.to,hashnode.com, hosts containingplainenglish).
Publisher matching rules: exact brand string for Dev.to, Hashnode, Medium, and freeCodeCamp; host regex for In Plain English (plainenglish in hostname).
| Scope | Value |
|---|---|
| Corpus window | 9 July – 29 October 2026 (developer-media UI captures) |
| Developer-media prompts | 332 |
| UI-family observations (deduplicated) | 508 |
| Observations with ≥1 cited source | 196 (38.6%) |
| Alignment panel (cited answers used below) | 191 |
| Total citation URLs (category) | 1,707 across 602 unique hosts |
Browse live aggregates in Developer Media on Explorer, including platform splits such as Claude and Perplexity.
What we found
1. Hashnode and Dev.to lead recommendations; citations lag
On 191 cited developer-media answers (latest observation per engine on each prompt):
| Outlet | Mentioned | Linked (matching host) | Both mention + link | Mentioned, not linked |
|---|---|---|---|---|
| Dev.to | 69.6% (133) | 26.7% (51) | 23.6% (45) | 46.1% (88) |
| Hashnode | 65.4% (125) | 5.2% (10) | 5.2% (10) | 60.2% (115) |
| In Plain English | 44.0% (84) | 41.4% (79) | 31.4% (60) | 12.6% (24) |
| Medium | 43.5% (83) | 11.0% (21) | 6.8% (13) | 36.6% (70) |
| freeCodeCamp | 32.5% (62) | 2.6% (5) | 2.1% (4) | 30.4% (58) |
Citation-to-mention ratio on this panel:
| Outlet | Ratio (cite % ÷ mention %) |
|---|---|
| In Plain English | 0.94× |
| Dev.to | 0.38× |
| Medium | 0.25× |
| Hashnode | 0.08× |
| freeCodeCamp | 0.08× |
Hashnode is the largest gap: it appears in roughly two of three sourced answers, yet hashnode.com is linked in roughly one of twenty.
2. Citation share and mention share invert for Plain English vs Hashnode
Across all 508 deduplicated UI observations in the category (sourced and unsourced):
| Outlet | Share of brand mentions | Share of citation URLs |
|---|---|---|
| Dev.to | 5.6% (#1) | 6.9% (#2) |
| Hashnode | 5.2% (#2) | 2.2% |
| In Plain English (brand) | 3.1% (#4) | — |
| plainenglish.io hosts combined | — | 11.5% (#1) |
resources.plainenglish.io alone is 9.5% of category citations — 3.0× the brand’s mention share. Hashnode’s mention share is 2.4× its citation share.
Host concentration also runs ahead of brand concentration: citation HHI 0.0196 vs brand mention HHI 0.0141 on the same observation set.
3. Search-heavy engines show the same pattern
Among cited answers only, Perplexity (n=80) and Claude (n=86) still mention Hashnode on 76–49% of answers while linking Hashnode on ≤5%:
| Engine | Dev.to mention / cite | Hashnode mention / cite | In Plain English mention / cite |
|---|---|---|---|
| Perplexity | 78.8% / 31.3% | 76.3% / 5.0% | 48.8% / 45.0% |
| Claude | 55.8% / 16.3% | 48.8% / 2.3% | 34.9% / 30.2% |
Perplexity is the closest on In Plain English alignment (45.0% cite vs 48.8% mention). Neither engine closes the Hashnode gap.
ChatGPT’s smaller cited subset (n=57 in category) shows the strongest citation concentration (host HHI 0.049) but still lists Hashnode far more often than it links hashnode.com.
4. Corpus-wide checks on the same prompts
On 196 observations with sources, 45.9% mention Dev.to while dev.to is absent from the link list. 12.8% mention In Plain English without a plainenglish host — the symmetric miss is smaller.
Why this is surprising
Developer-media answers look like a single ranked list of “best” outlets. When engines attach sources, many answers split the signal: training-familiar names (Hashnode, Dev.to, freeCodeCamp) fill the recommendation bullets, while retrieval-selected URLs cluster on a smaller set of publishers — especially Plain English subdomains.
A brand tracking only cited domains in this category would conclude In Plain English dominates distribution; a brand tracking mentions would overweight Hashnode and Dev.to. Neither view alone matches what buyers see in full answers.
This is distinct from cross-engine #1 disagreement on the same prompts (documented separately): here, within one answer, mentions and links disagree.
Limitations
- Only 38.6% of developer-media UI observations include sources; alignment metrics apply to that minority, not all answers.
- Brand extraction depends on list ordering and string matching; host matching uses hostname rules and may miss redirects or regional domains.
- Category sample is 332 prompts — substantial for a vertical slice, not a global corpus claim.
- API batch captures (excluded when a UI capture exists) may behave differently; this report focuses on consumer UI families.
Methodology
Platforms: ChatGPT, Gemini, Claude, Perplexity, and Grok UI captures in the research corpus.
Prompts: Active buyer-style developer-media queries with no customer website scope — the same public research set surfaced in Explorer.
Deduplication: Latest observation per prompt per platform family; UI platform keys preferred over API aliases when both exist.
Mentions: Distinct brand strings in the extracted ordered list from each answer.
Citations: Unique hostnames from sources URLs per observation; Plain English tallies any host containing plainenglish.
Exclusions: Observations outside the five UI families; prompts without developer-media subvertical tagging.
Date window: Earliest to latest developer-media UI observation in the research corpus through the November 2026 export (2026-07-09 – 2026-10-29 for UI rows in this category).
Conclusion
For developer media, recommendation lists and citation lists measure different things. Hashnode and Dev.to win mindshare in extracted brand lists; Plain English wins link share when engines attach sources. Visibility programs that treat citations as a proxy for “who AI recommends” will mis-rank outlets in this category — unless they score mentions and citations separately on the same prompts.