Three lists we can defend line by line.
No composite score, no opaque index. Every row here is a citation count over the runs we could actually measure, next to the search rank that source entered at, and it links to the query it came from. If a number cannot be traced back to a raw run, it is not on this page.
01 · Most cited
Ordered by the share of measurable runs that cited the source. Ties are broken by how many runs we could measure, then by search rank — never by anything we were paid. Showing the top 12 of the 22 sources that were cited at least once; the rest are in the directory.
aigradetool.com
Free AI Visibility Score for Google AI Overviews, ChatGPT
aivisibilityranking.com
AI Visibility Ranking 2026 - Tool Comparison & Rating
llmstxtgenerator.dev
LLMs.txt Generator — Create AI-Ready Website Files for Free
ppt.ai
PPT AI - AI Presentation Maker, Fast and Easy | PPT.AI
csdn.net
2026年指南：robots.txt如何放行AI爬虫抓取
citegrade.com
Citegrade — Get Your Website Cited by ChatGPT & AI Search
blog.csdn.net
AI crawler robots.txt 2026 gptbot claudebot perplexity
genppt.com
AI Presentation Generator | GenPPT
llmstxtpro.com
Free LLMs.txt Generator & Validator – Free LLMs.txt
crawlreadyai.com
How to Create llms.txt (or llm.txt) Step by Step
aipowerpointmaker.com
AI PowerPoint Maker, PPT Generator, Slides AI
wps.com
AI PPT Maker – Professional Presentations Made Simple
02 · Ranked first, cited zero times
These pages sat at the top of the search results we built each pool from, and the model cited someone else every single time we asked. A high rank is not a citation — this list exists because the two numbers are constantly confused.
baike.baidu.com
llmstxt-generator - 百度百科
mobileproxy.space
llms.txt 教程：如何把你的内容交给 AI 搜索引擎
aisearchvisibility.com
AI Search Visibility Tool - Compare Top 10 AI Visibility Tools & Trackers
aipresentationmakers.com
AI Presentation Makers | Reviews and Recommendations
microsoft.com
AI PowerPoint 生成器 | Microsoft PowerPoint
03 · Which queries are still open
How much the cited set changes between runs of the same query. Low overlap means no page owns the answer yet, which makes that query the one worth working on. This is the most useful list on this page if you are deciding where to spend effort.
best ai ppt maker — run-to-run overlap 0.472 (range 0.250–0.667). That is not settled. Who gets cited changes from run to run, which means the slot is still open — this is the kind of query worth working on.
llms.txt generator — run-to-run overlap 0.544 (range 0.250–1.000). That is not settled. Who gets cited changes from run to run, which means the slot is still open — this is the kind of query worth working on.
best ai visibility tracking tool — run-to-run overlap 0.544 (range 0.250–1.000). That is not settled. Who gets cited changes from run to run, which means the slot is still open — this is the kind of query worth working on.
how to get cited by chatgpt — run-to-run overlap 0.628 (range 0.250–1.000). That is partly settled — there is a stable core, and room around it.
robots.txt allow ai crawler gptbot — run-to-run overlap 0.700 (range 0.250–1.000). That is partly settled — there is a stable core, and room around it.
How these lists are made
A position cannot be bought. Every list is ordered by a measurement. There is no price at which payment changes a position, and sponsored placements — when we have any — are labelled, carry rel=sponsored, and never enter a list ordered by citation.
Every row traces back to a raw run. The citation count is a numerator over a denominator of runs we could actually read, and the denominator differs per query because some runs failed to parse. Those runs are excluded from both sides, not counted as zeros.
A zero is published, not hidden. Sources that were never cited appear on this page with a zero, and the directory shows the same number. A zero is a complete result, and it is usually the most actionable one.
Measured 2026-10-16 · gpt-oss-20b via text.pollinations.ai (alias "openai") · read the page body · 28 of 30 runs usable · denominators differ per query
These are correlations from one small experiment, not causal claims. n is small, so treat every ordering here as a hypothesis with a receipt attached.
Every listing in the directory carries its own citation count, and the zeros are there too.