Answers
Original research on how AI picks what to recommend.
We are early, so we would rather show the method than a cherry-picked number. Everything here comes from our own runs, and every figure carries the source it came from. Our own measurement so far covers 901 citations across 60 answers (source: our own runs), of which 91.5 percent pointed off-site, spread over 82 distinct hosts.
Short answer
What is on the Citon blog?
Original research on how AI assistants choose which tool to recommend. Everything here comes from our own runs, and every figure carries its source.
AI Overview Optimization: What Actually Changes
Google ranks first for "AI Overview optimization" and says no special optimization exists. What the docs state, what the studies measured, what nobody tested.
42 min
GPTBot, OAI-SearchBot, and Which Crawler Actually Decides Whether ChatGPT Can Cite You
OpenAI runs four crawlers and only one of them decides whether ChatGPT search can cite you. Which one it is, and why the popular warning runs backwards.
46 min
Does llms.txt actually move AI citations, and how would you tell
Every page ranking for this term explains what llms.txt is. None answers whether it changes anything. What 40 sites deploy, and the test that would tell.
43 min
Prompt Tracking, How Many Prompts and How Many Repeats Before a Change Is Real
Every page ranking for prompt tracking says how to pick prompts. None says how many, or how many repeats, before a weekly change stops being noise.
44 min
How to Track AI Visibility, and How to Read What It Tells You
The trackers solved collection and left interpretation alone. The record schema, the sampling trade, the null distribution, and whether your design could see a win.
41 min
LLM SEO, what it actually is and what the work looks like
Google says its AI features run on the same ranking systems as Search, so the tactics are familiar. What changed is how you tell whether the work landed.
46 min
AirOps Alternatives: Content-AEO vs Pure Measurement
Nine real AirOps alternatives, split into content production and pure measurement, plus the buyer split most vendor comparison pages miss entirely.
25 min
Best AI Visibility Tools (and Where a Tool Alone Runs Out)
Seven AI-visibility tools compared on what they actually measure and what they cost, plus the one question none of them can answer on their own.
24 min
Profound Alternatives: What Teams Actually Switch To (and Why)
Five real alternatives to Profound compared on what they track and what they cost, built from the actual switching threads where teams explain why they left.
23 min
Best AI SEO Agency for Developer Tools and API Companies
A five-question checklist for vetting an AI SEO agency, built from what real developer-tool buyers report going wrong, not a ranked vendor list.
24 min
Does Google index Reddit comments? What actually counts
Reddit's robots.txt disallows every crawler, yet Google shows Reddit with full descriptions. Here is what is actually indexed, measured rather than assumed.
25 min
AI Citation Decay, Half-Life, and When a Drop Is Real
A competitor platform now tracks how long an AI citation survives. We checked the idea against our own volatility data. A single drop is not proof of decay.
23 min
Does Reddit Marketing Move AI API Recommendations?
Reddit discloses a 21% AI Overview citation share and 46.7% on Perplexity. Four agencies pitch Reddit marketing as the fix, none publish a held-out test.
23 min
Profound Pricing: What an AI Visibility Tool Costs
Profound publishes two tiers and hides the third. What Starter and Growth gate, and what Enterprise costs per brand per country, per one practitioner's report.
23 min
What Predicts an AI Citation Is Not What Proves Yours
DiscoveredLabs measured 2 million AI citations and found what correlates with getting cited. Correlation across pages you don't own isn't proof for yours.
22 min
AI Visibility Tools Show a Score, Not Whether It Was You
Ten guides rank the best AI visibility trackers. None of them tells you whether the score moving was your AEO spend or the model changing its mind on its own.
23 min
Reddit is cited by AI. That is not a reason to buy upvotes.
Published Reddit citation shares run from 2% to 46.7%, and one study puts Reddit at 67.8% of every URL ChatGPT retrieves and then declines to cite.
23 min
Reddit Blocks 23M Spam Views a Day. A Citation Signal.
Reddit published its spam enforcement numbers for users and moderators. Read from the AI-citation side, it is an argument about which sources deserve quoting.
26 min
What Is Answer Engine Optimization? How AI Picks Sources
What answer engine optimization means, how AI assistants decide what to cite, and what our own measurement work says about proving it caused a citation gain.
23 min
Why a single AI visibility score is noise
Twelve money queries, five identical repeats, one model, one day. Seven changed their answer, which breaks every before-and-after published in this category.
24 min
Frequently asked questions
- What does the Citon blog cover?
- How AI assistants choose the sources they cite, what the evidence actually supports about that, and what has to be true of a measurement before a change in citations can be credited to the work rather than to the churn.
- Is the research here original, or a summary of other people's?
- Original. Posts draw on our own runs, and where a figure comes from somewhere else it is attributed at the point of use rather than in a footnote. Our own measurement so far covers 901 citations across 60 answers, of which 91.5 percent pointed off-site, spread over 82 distinct hosts.
- How often is it updated?
- New posts publish when a measurement is finished rather than on a fixed calendar. Existing posts are revised in place when a rerun changes what the evidence supports, and the change date moves with the content.
- Can I reuse a chart or a figure from a post?
- Yes, with a link back to the post it came from. Please take the number and its stated basis together, since most figures here are only meaningful alongside the sample and the window they were measured over.
Every answer
The full index, by category.
Three categories, and every post in each one. The card grid above shows what is newest; this shows everything, so nothing published here is reachable only from a sitemap.
3 categories
21 posts
How assistants choose which tool to name, which sources they pull from when your buyers ask, and what actually moves a citation once it is measured.
- AI Overview Optimization: What Actually Changes42 min
- GPTBot, OAI-SearchBot, and Which Crawler Actually Decides Whether ChatGPT Can Cite You46 min
- Schema Markup for AI Search: How to Isolate the Variable (2026)80 min
- Does llms.txt actually move AI citations, and how would you tell43 min
- LLM SEO, what it actually is and what the work looks like46 min
- Best AI SEO Agency for Developer Tools and API Companies24 min
- Does Google index Reddit comments? What actually counts25 min
- AI Citation Decay, Half-Life, and When a Drop Is Real23 min
- What Predicts an AI Citation Is Not What Proves Yours22 min
- AI Visibility Tools Show a Score, Not Whether It Was You23 min
- Reddit is cited by AI. That is not a reason to buy upvotes.23 min
- Reddit Blocks 23M Spam Views a Day. A Citation Signal.26 min
- What Is Answer Engine Optimization? How AI Picks Sources23 min
Whether any of this can be measured at all, what a number has to survive before it means anything, and why a single visibility score is noise.
- Prompt Tracking, How Many Prompts and How Many Repeats Before a Change Is Real44 min
- How to Track AI Visibility, and How to Read What It Tells You41 min
- AirOps Alternatives: Content-AEO vs Pure Measurement25 min
- Best AI Visibility Tools (and Where a Tool Alone Runs Out)24 min
- Profound Alternatives: What Teams Actually Switch To (and Why)23 min
- Profound Pricing: What an AI Visibility Tool Costs23 min
- Why a single AI visibility score is noise24 min
The surfaces a developer actually evaluates from, why docs and community threads decide the answer, and where API buyers look before they ever visit you.
Start here
See what your buyers are actually being told.
Four questions, run against live AI answers, walked through on a call. No deck.
Ask for the Gap Report