About Experience Learn Insights AI Tools Contact Book a call

AI Visibility

I earned 40,000 AI citations. Most AI SEO advice had nothing to do with it.

Five ordinary websites. Zero to 42,000 AI citations in six months. Most of the standard AI SEO checklist had nothing to do with it. Here is what did, in plain English.

"AI visibility" is the new thing everyone wants to sell you, and the pitch is always a checklist: schema markup, longer guides, an llms.txt file. I did not have to guess whether it works. AI engines have cited my five websites more than 42,000 times, and I have the receipts for what earned it. Most of the checklist is not on the list.

Plain English first: an AI citation is when ChatGPT, Copilot, or Perplexity answers a question and names your website as the source. It is the new page one. My sites are nothing special. No famous brand, no ad budget, one operator.

The scoreboard: zero to 42,000 in six months

0 10K 20K 30K 40K Feb Mar Apr May Jun Jul first citation · Feb 22 the ramp begins · Apr demand window · +8,859 in one week 42,077
Cumulative AI citations across the portfolio, weekly, January 17 to July 27, 2026. Source: Bing Webmaster Tools, AI Performance report, all sites summed. The jump in July is a game-update demand window on the gaming site. Hover or tap the line for weekly numbers.
See the weekly numbers as a table
Week endingCitations that weekRunning total
42,000+AI citations in six months
5ordinary websites, five niches
58pages doing the winning
8,800+from a single page

One counter, Bing's free AI Performance report, and it only sees Microsoft's AI surfaces. If anything, it undercounts.

The growth is compounding

The month view is the part that still surprises me. June roughly tripled May. July more than tripled June, with four days still uncounted. That is what happens when you publish what AI has to look up.

15 71 3,202 2,809 8,372 27,608 Feb Mar Apr May Jun Jul
AI citations earned each month, all sites combined. July is counted only through July 27.

Now the claims. Each headline below is something the AI SEO industry repeats. The verdict is what my data says.

"AI only recommends big brands" Myth

This is the one that keeps small businesses from trying, and it failed hardest. My sites are small, niche, and unknown, and they were cited 40,000+ times. In my live tests, tiny sites beat major brands whenever they had the more specific, current answer. A contractor with the only current permit-cost page in the county can out-cite a national chain.

"You need schema markup to show up in AI answers" Myth

My most-cited page has 8,800+ citations and zero schema markup. Across my sites the pattern runs backwards: the pages AI never cites follow the formatting checklist more than my winners do. Keep schema for regular SEO. Do not pay for it as an AI strategy.

Pages AI never cited 99% Pages AI cites 66%
Share of pages carrying the "format it for AI" checklist, from an 875-page audit of my own sites. The never-cited pages follow it more.

"Longer, more detailed content wins" Myth

I tested 751 pages on one identical template, where only topic and depth varied. Word count made no detectable difference, and 84 percent of my longest, deepest pages earned exactly zero. Write enough to answer the question, then stop.

"llms.txt is the new secret handshake" Myth

My sites have polished llms.txt files, including pages with zero citations. Nothing in 40,000 citations suggests the file earned any of them. Cheap, harmless, not a strategy. If a pitch leads with llms.txt, that tells you how thin the rest of it is.

"AI SEO is a brand-new discipline" Myth

About 80 percent of it is the oldest advice on the internet: rank in regular search. When an AI searches, it mostly cites what its search backend returns near the top. My biggest page ranks near the top of Google for its main query, and that rank, not any AI trick, is why engines keep grabbing it. The new part is small: be readable without JavaScript, and give the machine one answer it can quote.

"Fresh, updated content gets picked first" Real

This one the industry gets right. My biggest winner updates weekly, and engines learned to trust it for current answers. When the game my stats site covers shipped a new season, I regenerated every page the same day. That week added 8,859 citations, the spike in the first chart. The page only you keep current is the page AI has to cite.

"Formatting decides what gets quoted" Half-true

Structure decides what gets quoted once you are in the game. It does not put you in the game. My best page earned its first 6,000 citations before I added any of it. Answer block up top, one clean table, a visible date. Then move on.

The rule nobody is selling: if the AI does not search, nobody gets cited

When someone asks an AI a question, the model first decides whether it already knows the answer. If it thinks it does, it never searches and cites no one. Not you, not your competitor, no one. That is why your "10 tips" post is invisible while your current price sheet forces the AI to go look.

Topics people actually search 47% cited Topics they don't 6%
Same template, same effort, roughly seven times the citation rate when the topic has real demand. From my 751-page test.

Publish what the model cannot already know. Everything else on this page comes after that decision.

The six things to actually do

This is the whole program. Every step is free or close to it.

  1. Open your scoreboard today. Bing Webmaster Tools, AI Performance report, free. Write your number down, even if it is zero. Mine was zero in February.
  2. Pick one topic AI cannot answer from memory. Current prices, local rates, what a new rule means in your trade. If the honest answer needs today's numbers, the AI has to search.
  3. Win the boring search result. Rank for the question, and make sure your key facts show up without JavaScript.
  4. Give the machine one clean thing to lift. Full answer in the first three sentences, one table, a visible date.
  5. Update on a calendar, and ping when you publish. IndexNow is free, and it is most of the difference between day four and day 64.
  6. Measure it monthly, like money. Export the Bing report. Track referrals from chatgpt.com, perplexity.ai, and copilot.com. Quarterly, ask ChatGPT your ten biggest customer questions and see who gets cited.
Brand-new site, pings on 4 days Established site 6 to 8 days No pings, weak search presence 64 days
Days from publishing to first AI citation, across my sites. Free publish pings are most of the difference.

Questions business owners actually ask

Does AI SEO actually work?

Parts of it. What worked across my 40,000+ citations: topics the AI has to search for, ranking in regular search, keeping pages current, and one clean liftable answer. What did not: schema markup, extra word count, and llms.txt. My most-cited page has 8,800+ citations and zero schema.

How do I get ChatGPT to recommend my business?

Publish the page only you can keep true: current prices, local rates, what changed this season. When the AI cannot answer from memory it searches, and it mostly cites what ranks near the top. In my tests, small sites beat big brands whenever they had the more specific, current answer.

Do I need to hire an AI SEO agency?

Not for the checklist most of them sell. In my data, topic choice and search presence decided the winners, and both are things you can work on directly. Measurement is free in Bing Webmaster Tools.

How long does it take to show up in AI answers?

New pages on my established site earned first citations in 6 to 8 days. A brand-new domain with publish pings: four days. A site without pings: 64 days. Crawl setup and search rank set the speed.

How do I measure AI visibility for free?

Verify your site in Bing Webmaster Tools and open the AI Performance report: citations per page, per day, and the queries behind them, free. Then track referrals from chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, and copilot.com in your analytics. That is a real scoreboard.

Where to go deeper

The step-by-step version of everything here, with the receipts attached, is the free playbook, How to get cited by AI. The research cuts live in the field notes: why demand beats depth and how long a first citation takes. The free GEO course covers the same ground in five short lessons.

And if you want an honest read on where your own site stands, which pages could earn citations and which are graveyard plots, book a clarity call. One conversation, no pitch.

Paul Takisaki

Paul Takisaki

Strategic Advisor on AI, Leadership & Growth. Former Verizon Associate Vice President and four-time President's Cabinet winner who turned around four major markets, including 19 consecutive months of YoY growth in the Pacific Northwest. Now running two AI-powered businesses solo and building the systems behind them.

Related field notes

More field notes