Same prompt, different answer. AI search platforms serve different models, different data, and different brand recommendations depending on whether you’re logged in, what you pay, and which platform you’re on. Here’s what that means for your visibility strategy.
Two people at the same company open ChatGPT and ask the same question. “What’s the best email marketing platform for a 50-person team?” One is on a free account, browsing without signing in. The other is on ChatGPT Pro, logged in, with memory turned on. The recommendations come back different. Different brands. Different ordering. Different justifications.
It is not a bug. It is the architecture.
AI search platforms don’t have one consistent voice. They have many, layered together: a model tier, a user state, a memory state, a browsing state. Every one of those layers can change which brands surface in an answer. And most brands tracking their AI visibility don’t realize they’re measuring just one of those layers.
The model tier changes the brain
Free ChatGPT, ChatGPT Plus, ChatGPT Pro, ChatGPT Team. Those aren’t pricing tiers for the same product. They unlock different underlying models. Free tier might be running a smaller, faster model. Paid tiers run larger, more capable models with deeper reasoning, longer context, and stronger brand recall.
The same is true at Gemini (free vs Advanced), Perplexity (free vs Pro), and Claude (free vs Pro vs Team). The bigger the model, the more nuanced its brand judgments tend to be. A smaller model might fall back on the three most-mentioned brands in its training data. A larger one can reason about category fit, user constraints, and edge cases.
If your brand is well-known but rarely cited in context, you might surface on a paid tier and disappear on free. If you’re a niche specialist, the opposite can happen.
Logged in vs logged out isn’t cosmetic
When a user is signed in, the model has access to context the anonymous user doesn’t: past conversations, stored preferences, and in some cases memory of previously mentioned brands. ChatGPT’s memory feature can lock in a brand preference after a single positive mention. Once it’s in memory, that brand surfaces more often in future answers.
This creates a feedback loop. A user who mentioned your brand once in passing six months ago may see your brand recommended every time they ask about your category. A new user with no history gets the unfiltered baseline.
For visibility tracking, both states matter. The logged-in state predicts retention. The logged-out state predicts discovery.
Live browsing changes everything
When a platform browses the web in real time, the answer it gives isn’t based solely on training data. It’s based on what’s ranking today, what’s being cited in current reviews, and what shows up in recent press. That’s a completely different signal from what’s baked into model weights from training.
The result: a brand can be invisible in the model’s pre-trained baseline and still be heavily recommended when the browsing layer kicks in. Or vice versa: a brand that dominated training data three years ago might lose ground to a fresher competitor whose recent press shows up in the live browse.
Whether browsing is enabled depends on the platform, the user’s settings, and sometimes the specific prompt. Some platforms route browsing on automatically when they detect a recommendation query. Others require the user to opt in.
Personalization layers on top
Beyond memory, several platforms quietly personalize. They use the user’s IP location, language settings, prior session data, and inferred interests to tilt recommendations toward what each individual is most likely to engage with.
A buyer in San Francisco asking about CRMs may get a different brand list than the same buyer asking from Austin. A user who has spent time in software-engineering threads might see developer-focused tools recommended over sales-focused ones.
You can’t see this personalization directly. But it’s there, and it means the “AI search results” your competitor sees are not the ones your prospect sees.
An AI visibility audit done on one logged-out anonymous browser is not an audit of how AI sees your brand. It is one slice of one user state on one platform.
What brands should actually do
Knowing this changes what AI visibility tracking has to look like.
Audit across states, not just one
The same prompt on free ChatGPT (logged out, no browsing) and ChatGPT Pro (logged in, browsing on) should be tracked as different surfaces. The AiR Dashboard runs every prompt across multiple tiers and states for exactly this reason.
Optimize for the consistent signals
Optimize for the signals that show up regardless of tier. Brands that appear in baseline training data, get cited in fresh editorial coverage, and get mentioned in recent reviews are the ones that survive the variance. Optimizing just for one layer is fragile.
Stop treating “AI search” as one entity
ChatGPT free is a different surface than ChatGPT Pro. Gemini’s anonymous tab is a different surface than Gemini Advanced with memory on. Track them separately, optimize for the patterns that hold across them.
The brand that wins AI visibility in 2026 is the brand that surfaces on every tier, every state, every platform. Not the one that maximized for a single test in a single browser window.
Want to see how your brand shows up across tiers and platforms? See how we approach Google & SEO, or get in touch and we’ll run the audit across every state.