{"id":42115,"date":"2026-08-20T07:27:07","date_gmt":"2026-08-20T05:27:07","guid":{"rendered":"https:\/\/www.graviton.at\/letterswaplibrary\/99-days-of-ai-answer-engine-responses-to-a-fixed-16-question-set-24882-scored-answers-from-openai-search-gemini-and-claude-plus-a-10-model-no-web-control-cc-by-4-0\/"},"modified":"2026-08-20T07:27:07","modified_gmt":"2026-08-20T05:27:07","slug":"99-days-of-ai-answer-engine-responses-to-a-fixed-16-question-set-24882-scored-answers-from-openai-search-gemini-and-claude-plus-a-10-model-no-web-control-cc-by-4-0","status":"publish","type":"post","link":"https:\/\/www.graviton.at\/letterswaplibrary\/99-days-of-ai-answer-engine-responses-to-a-fixed-16-question-set-24882-scored-answers-from-openai-search-gemini-and-claude-plus-a-10-model-no-web-control-cc-by-4-0\/","title":{"rendered":"99 Days Of AI Answer Engine Responses To A Fixed 16-question Set: 24,882 Scored Answers From OpenAI Search, Gemini And Claude, Plus A 10-model No-web Control (CC BY 4.0)"},"content":{"rendered":"<p><!-- SC_OFF --><\/p>\n<div class=\"md\">\n<p>Disclosure: this is my own dataset. I built and run the measurement, and the subject of the measurement is my own pen name, so treat the topic with that in mind. The data itself is mechanical: same 16 questions, every day, three answer engines with web access, answers scored with a fixed rubric.<\/p>\n<p>Dataset (Hugging Face, CC BY 4.0): <a href=\"https:\/\/huggingface.co\/datasets\/marintkael\/ai-citation-fidelity\">https:\/\/huggingface.co\/datasets\/marintkael\/ai-citation-fidelity<\/a><\/p>\n<p>Five configs:<\/p>\n<ul>\n<li>default: 24,882 scored answers including the no-web control channel<\/li>\n<li>claude_web: 3,279 answers from a separate Claude web search panel<\/li>\n<li>questions: the 16 questions with category and channel<\/li>\n<li>data_gaps: register of measurement gaps (provider outages, method changes), because a gap is not a zero<\/li>\n<li>daily_channels: daily time series split into direct, long tail and discovery channels<\/li>\n<\/ul>\n<p>Collection window 2026-05-13 to 2026-08-19 (99 measurement days). The no-web control is 6,416 blind answers from 10 models without web access, useful for separating retrieval effects from training data. Scoring code and figure scripts: <a href=\"https:\/\/github.com\/marintkael\/marin-research-tools\/tree\/main\/reports\/03-found-not-recommended\">https:\/\/github.com\/marintkael\/marin-research-tools\/tree\/main\/reports\/03-found-not-recommended<\/a><\/p>\n<p>Written report describing method and findings, if you want context before touching the parquet: <a href=\"https:\/\/doi.org\/10.5281\/zenodo.22015495\">https:\/\/doi.org\/10.5281\/zenodo.22015495<\/a><\/p>\n<\/div>\n<p><!-- SC_ON -->   submitted by   <a href=\"https:\/\/www.reddit.com\/user\/marintkael\"> \/u\/marintkael <\/a> <br \/> <span><a href=\"https:\/\/www.reddit.com\/r\/datasets\/comments\/1vta1jh\/99_days_of_ai_answer_engine_responses_to_a_fixed\/\">[link]<\/a><\/span>   <span><a href=\"https:\/\/www.reddit.com\/r\/datasets\/comments\/1vta1jh\/99_days_of_ai_answer_engine_responses_to_a_fixed\/\">[comments]<\/a><\/span><\/p><div class='watch-action'><div class='watch-position align-right'><div class='action-like'><a class='lbg-style1 like-42115 jlk' href='javascript:void(0)' data-task='like' data-post_id='42115' data-nonce='9de69db8d5' rel='nofollow'><img class='wti-pixel' src='https:\/\/www.graviton.at\/letterswaplibrary\/wp-content\/plugins\/wti-like-post\/images\/pixel.gif' title='Like' \/><span class='lc-42115 lc'>0<\/span><\/a><\/div><\/div> <div class='status-42115 status align-right'><\/div><\/div><div class='wti-clear'><\/div>","protected":false},"excerpt":{"rendered":"<p>Disclosure: this is my own dataset. I built and run the measurement, and the subject of the&#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[85],"tags":[],"class_list":["post-42115","post","type-post","status-publish","format-standard","hentry","category-datatards","wpcat-85-id"],"_links":{"self":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts\/42115","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/comments?post=42115"}],"version-history":[{"count":0,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/posts\/42115\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/media?parent=42115"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/categories?post=42115"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.graviton.at\/letterswaplibrary\/wp-json\/wp\/v2\/tags?post=42115"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}