From Click to Chat: A Comparative, Bilingual Study of Autism Related Information on Websites and General Purpose AI Chatbots (Preprint)
Bibliographic record
Abstract
Background: Parents increasingly consult the internet, both websites and, more recently, artificial intelligence chatbots, for information on autism spectrum disorder (ASD). However, the comparative quality of these two source types, especially across languages, remains underexplored. Objective: This study aimed to assess the completeness and accuracy of ASD information delivered by websites and 5 popular artificial intelligence chatbots and determine whether performance differs between English and Romanian content. Methods: In a cross-sectional design, 25 English-language and 25 Romanian-language websites and the responses of ChatGPT, Gemini, Claude, Copilot, and DeepSeek were evaluated. Content was benchmarked against a 24-item checklist, yielding completeness and accuracy scores. Chatbots were tested in 2 scenarios: a single broad query (A) and 24 item-specific queries (B). Results: Websites achieved higher completeness in English than in Romanian (6.9 vs 5.1; P=.007) and marginally higher accuracy (6.9 vs 6.1; P=.045). In scenario A, chatbot completeness (English: 5.3 vs Romanian: 6.2; P=.15) and accuracy (English: 6.0 vs Romanian: 5.6; P=.32) did not show significant differences by language. In the single-query scenario, websites showed higher accuracy than chatbots in both English (6.9 vs 6.0; P=.19) and Romanian (6.1 vs 5.6; P=.53), with neither difference reaching statistical significance. Conversely, item-specific questioning favored chatbots, which yielded higher accuracy scores than websites in English (8.3 vs 6.9; P=.053) and Romanian (8.0 vs 6.1; P=.007). Accuracy scores improved significantly from the single-query to the item-specific scenario (English: 6.0 vs 8.3; P=.002; Romanian: 5.6 vs 8.0; P=.001). While an initial analysis suggested variation in performance between chatbots (repeated measures ANOVA; P=.005), pairwise differences between individual models did not remain significant after adjustment for multiple testing. Conclusions: This exploratory study indicates that the quality of online ASD information varies by language and source context. English-language websites are more complete than Romanian-language websites. Among chatbots, targeted questioning yields more accurate answers than single broad queries in both languages. The findings should be interpreted cautiously due to the temporal gap between website and chatbot data collection.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.010 | 0.028 |
| Meta-epidemiology (narrow) | 0.000 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.002 | 0.002 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".