AI Search Optimization Services
Being quoted by ChatGPT isn't the same as being recommended by it. Only one of those sends you customers. Here's how I work out which one you're getting.
Updated July 2026 · Sources cited at the bottom of the pageMost AI Search Work Fixes the Wrong Problem
The whole field is selling page formatting. Chunk your text, add schema, hit a readability score, publish a “best in category” list.
Those are on-page levers. The problem they're aimed at lives off your site.
Lily Ray put it plainly after testing it: Google appears to have decoupled what it cites from who it recommends. The recommendation is the part that matters, by an order of magnitude.
So the question isn't whether a model can read your page. It's whether the rest of the web gives it a reason to name you.
The Four States Your Business Can Be In
Every AI answer has two layers. The body names brands. The footnotes list sources.
They're not the same list. That gap puts you in one of four states.
Invisible
Not named, not cited. The model doesn't know you exist.
It's the honest bad state. Nothing is being taken from you.
Harvested
Cited but not named. Your page supplies the answer and a competitor gets the credit.
This is the worst one commercially, and almost nobody spots it on their own. You're doing the work and someone else collects.
Known
Named but not cited. The model trusts your brand from memory and sends the click elsewhere.
That's a plumbing problem. Give it something of yours to point at.
Recommended
Named and cited. That's the one that pays, and it's what the work builds toward.
What I Actually Do
Three things, in this order.
1. The diagnosis
I run the questions your buyers actually ask through the models they actually use. Then I count two things separately: how often you're named, and how often you're cited.
That tells us which of the four states you're in. Everything after this depends on that answer.
2. The crawlability floor
If a model can't fetch and read the page, nothing else matters. Blocked bots, broken rendering, dead links, slow servers.
This part is real and it's mine to fix. It won't make you recommended on its own, and I won't pretend it will.
3. The off-site work
The recommendation axis moves when other sites talk about you and link to you. That's the slow part, and it's the part that actually shifts the outcome.
Real coverage, real citations from people who aren't you, and a first-page Google rank. Ray's data is blunt about that last one: ranking number one is what gets you the citations.
Want to know which of the four states you're in? Tell me what you sell and I'll tell you straight.
Ask me directly →What I Won't Sell You
Some of this field is selling things that don't work. Here's what I've stopped recommending, and why.
An llms.txt file
Two separate studies looked for an effect and found none. SE Ranking checked around 300,000 domains and found the file didn't make a domain more likely to be cited.
ALLMO went at it from the other end, looking at 94,614 cited URLs across 11,867 AI answers. Exactly one of those citations was an llms.txt page.
Two unrelated methods, same answer. I'll still add the file if you want it, but I won't charge you for it or call it a strategy.
Schema markup as a recommendation lever
Schema helps machines read a page. It doesn't make a model recommend a brand.
I'll fix your markup because it's cheap and it's part of the floor. It isn't the mechanism.
Listicles that name you the best
Writing your own “best in category” page has worked on language models, which is exactly why it's getting cleaned up. Ray tracked companies that scaled these into heavy declines starting January 2026.
At volume, near-identical comparison pages meet Google's existing definition of scaled content abuse. I don't build them.
How It's Priced
No custom quotes, no discovery-call theatre. The prices are on the page.
Start with the Deep Dive Audit at $1,497 if you want the full diagnosis first. Five findings or it's free.
If you'd rather just start the work, that's the Growth Retainer at $2,000/mo. Three needle movers ship every month or that month is free.
The Authority Retainer at $3,997/mo adds the off-site citation work on top. That's the one built for the recommendation axis.
Rankings take 8 to 12 weeks. Three months minimum, then cancel any time. Here's how the three fit together.
Two more pages if you're still reading rather than buying. The GEO tools I actually use, and my answer to whether AI is replacing SEO.
Questions I Get First
Is AI search optimization the same as SEO?
Mostly, yes. The models pull from search results, so a page that can't rank usually can't get cited either.
What's new is the second axis. You now have to be named, not just found.
How do I know if I'm being harvested?
Ask the models the questions your buyers ask, then check the footnotes against the answer text.
If your domain is in the sources and your name isn't in the answer, that's it. That's the state.
Do you guarantee I'll get recommended?
No. Anyone who does is guessing.
What I guarantee is the work: three needle movers a month or that month is free.
Which AI models do you check?
ChatGPT, Claude, Perplexity and Gemini. Those are the ones your buyers use.
Can you do this alongside my current SEO?
Yes, and usually that's cleaner. I'd rather fix the crawl floor and the off-site work than rewrite pages that already rank.
Where This Page's Claims Come From
- Lily Ray, ppc.land interview and lilyraynyc.substack.com, on citation and recommendation being decoupled
- Lily Ray, BuzzStream podcast with Vince Nero, 18 June 2026 · buzzstream.com/blog/lily-ray-podcast/
- Edward Sturm, edwardsturm.com, on LLMs sourcing from search engines
- SE Ranking, “LLMs.txt: Why Brands Rely On It and Why It Doesn't Work”, Yulia Deda, 7 Nov 2025 · seranking.com/blog/llms-txt/
- ALLMO, “LLMs.txt for AI Search Report 2026”, 23 Jan 2026 · allmo.ai/articles/llms-txt
Want to know which state you're in?
Tell me what you sell and who you compete with. I'll check the models and tell you honestly if I can help.