Why does ChatGPT recommend my competitor instead of me?
Recognition is not selection. A system can know your company exists and still decide that someone else fits the question better.
Founder and legibility strategist
AI knows your company.But would it choose you?
People are already asking ChatGPT, Claude, Gemini and Perplexity who to trust, hire and buy from. I measure what those systems understand about you, who they choose instead, why, and what can actually be changed.
Creator of the Legibility SprintControlled measurement, not visibility theatre.
Cross-model observationChatGPT, Claude, Gemini and Perplexity kept separate by surface.
Evidence limits statedNo guaranteed rankings and no invented causal claims.
The questions people actually ask
If any of these sound familiar, this is the problem I work on.
Why does ChatGPT know us but rarely mention us?
The failure can sit in retrieval, category fit, evidence, corroboration, or the decision criteria implied by the prompt.
How do I get my company recommended by AI?
There is no legitimate switch that buys a recommendation. What can be measured is the chain from retrieval to understanding, evidence, comparison and selection.
Is this just SEO for ChatGPT?
No. Search visibility is one input. Recommendation is a decision problem.
Does my website change what AI says about me?
Sometimes. But changing pages before diagnosing the failure can make the signal noisier rather than stronger.
Do I need Wikipedia?
Not automatically. Authority is the evidence environment a system can retrieve and reconcile, not one magic page.
Being known is not the same as being chosen.
Most companies are still optimizing for recognition: can the system identify us, cite us, summarize us?
The harder question is selection. When a user asks for the best fit under real constraints, do you enter the shortlist at all?
Recognition asks: does the system know you?
Selection asks: does the system choose you?
The method
I do not guess what AI thinks. I run the comparison.
Measure
Run controlled prompts across genuine consumer AI surfaces and preserve the raw answers.
Diagnose
Separate recognition, retrieval, evidence, corroboration and selection failures.
Intervene
Change only the public evidence the diagnosis actually implicates.
Measure again
Repeat matched conditions and inspect controls before calling a movement meaningful.
No guaranteed rankings. No invented AI visibility percentage. No claim of causation from one changed answer.
Legibility Sprint
What does AI currently think your company is, and who does it choose instead?
A controlled baseline across major AI systems, followed by a diagnosis of where the evidence chain breaks and a prioritized intervention plan. Raw observations stay visible. Interpretation sits on top of them.
- What each system currently understands about you
- Where you enter or disappear from relevant shortlists
- Which competitors or experts get selected instead
- Which sources and claims are carrying the answer
- What is actually worth changing first
- A repeatable T1 measurement after intervention
About
I got interested in this because I kept building systems that choose.
I opened my first business while I was still in medical school. Later I worked in cosmetology, built technology products, and architected SKINBOT, an AI decision layer used in live beauty-retail pilots.
The recurring problem was never just whether a model could answer. It was what evidence the system used, how it resolved a messy real-world category, and what made one option survive the decision while another disappeared.
That became my independent work on AI legibility and, more broadly, machine-mediated choice. I also founded getmai.ai, an AI engineering company.
Recognition is not selection. That gap is the work.
Read my storySelected research and essays
Start with the questions, then go deeper.
Questions, answered
What people usually ask before they care about the terminology.
Why does ChatGPT recommend competitors and not us?
Because knowing that your company exists is different from selecting it for a specific request. The system may retrieve stronger evidence for a competitor, categorize you differently, or infer that another option better fits the user's constraints.
Can I influence what ChatGPT says about my company?
You can improve the public evidence environment that AI systems read, but you cannot legitimately control a model's answer. The useful work is to diagnose what is wrong, change the relevant evidence, and measure again under comparable conditions.
Is AI legibility the same as GEO or AI SEO?
No. GEO and AI SEO usually focus on discoverability or citation. AI legibility includes those layers but continues into understanding, corroboration, comparison and selection.
Can you guarantee that ChatGPT will recommend me?
No. Nobody can credibly guarantee a consumer AI recommendation. What can be guaranteed is a controlled process: measure, diagnose, intervene and verify.
Why do ChatGPT, Claude, Gemini and Perplexity give different answers?
They use different models, retrieval systems, product rules, context and update cycles. A serious baseline preserves results by surface instead of averaging them into one answer.
Start with a baseline
Find out where the decision breaks.
If AI systems know your name but do not select you when the question matters, measure the current state before changing anything.