Can I simulate what AI search would say about my brand?

Any AI search simulator works the same way. It sends a question to a model, sometimes with web search switched on, and returns the answer. That answer is real output. It just comes from a configuration you chose, not the one your buyer is using.

The consumer apps add three layers a simulator usually can't copy:

• The model. Each app routes between models, and they don't share one knowledge cutoff.
• The retrieval decision. The app decides per question whether to search the web at all.
• The query. The app rewrites your buyer's words into its own search queries before it looks anything up.

So treat a simulator as a diagnostic, not a forecast. It tells you what a model can say about you given certain inputs. That's enough to find gaps in your facts. It isn't enough to promise what ChatGPT will tell a specific buyer tomorrow.

Models and knowledge cutoffs

As of October 2026, OpenAI documents a Feb 16, 2026 knowledge cutoff for its GPT-5.6 models and an Aug 31, 2025 cutoff for the Instant path (chat-latest). "ChatGPT" is not one knowledge state. Google says AI Overviews and AI Mode "may use different models and techniques", so even two Google surfaces can disagree.

Whether the engine searches

Retrieval is discretionary on every major platform. ChatGPT searches at the model's discretion. Claude.ai search is a user toggle. AI Overviews often don't trigger. Many simulators call an API with search off, which shows only what the model remembers from training.

Query rewriting

OpenAI states that ChatGPT search "typically rewrites your query into one or more targeted queries". Google may fan out into several related searches, and the Gemini API generates its own queries. The words you type into a simulator are not the queries that hit the index.

Personal context

Logged-in users bring memory and custom instructions. OpenAI notes that saved memories can shape how ChatGPT rewrites a search query. That's why logged-in and logged-out tests return different answers.

A simulator earns its place when you use it to find specific, fixable problems.

• Missing facts. If the answer leaves out your pricing, your main product or your location, check whether that fact is stated plainly on a page an engine can fetch.
• Wrong or stale facts. An old product name or a closed office is a concrete thing to trace. Knowing whether it's a hallucination or an outdated answer tells you where to look.
• Search on versus search off. Running the same question both ways separates training knowledge from live retrieval. A search on vs off test makes that comparison quick.
• Rendering gaps. Vercel found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them. If a fact only appears after JavaScript runs, a simulation with search on may never see it.

One free option is the AI Search Simulator from 99Visibility, which shows how engines pick sources for a query. This site's guide to the free AI search simulator explains how to read its output.

Nobody outside the engines can see the real retrieval queries, the full candidate list, or whether a given consumer answer used live search. A July 2026 critical survey of 45 GEO studies (arXiv 2607.14035) concluded that no reviewed technique shows "a stable, longitudinal, cross-platform causal effect on organic discoverability".

That has two consequences for simulation:

• A single score blends unlike things. Any number that averages answers across engines mixes answers given with and without retrieval. Read results per engine.
• Citation is not the same as use. OpenAI's API reports the sources a model consulted separately from the ones it cited, and the consulted list is often longer. A simulator that only counts citations undercounts influence.

What you control is narrower: whether your facts are stated clearly, whether crawlers can reach them, and whether your own pages agree with each other.

Say you run a dental clinic in Austin and want to know what AI says about your hours and services. A useful routine looks like this:

1. Write five to ten questions a real patient would ask, in different phrasings.
2. Run each one logged out, with memory off, on each engine you care about.
3. Run the same questions with search on and search off where the tool allows it.
4. Record each specific fact that is missing or wrong, with the date and engine.
5. Check that the correct fact appears in the raw HTML of a page AI crawlers can reach, and that your robots.txt and firewall let retrieval bots such as OAI-SearchBot in.
6. Fix contradictions on your own site, then rerun the same questions later.

Repeat after major model launches, because a new model release can change brand answers. For a wider method, see how to find out what AI says about your company.

Can I simulate exactly what ChatGPT will tell a customer? No. ChatGPT routes between models with different knowledge cutoffs, decides per question whether to search, and rewrites the question before searching. Logged-in users also bring memory and custom instructions. A simulation can match one of those configurations, but you can't know which one a given customer gets. Treat results as a sample that reveals missing or wrong facts, not as a transcript of a future conversation.

Why does the same question give different answers each time? Model output varies between runs, and the retrieval step adds more variation. The engine may search on one run and not the next, or pull a different set of pages. Over time, model updates and index changes shift answers further. Run each question several times and look for facts that are consistently missing or wrong, rather than reacting to a single answer.

Does a simulator show Google AI Overviews? Usually not directly. Most simulators call a model API, and the Gemini API with search grounding is a different path from AI Overviews or AI Mode. Google says those two surfaces may use different models and techniques, and AI Overviews don't appear for every query. If a tool claims to show AI Overviews, check whether it reads Google's results page or generates its own answer.

If a simulator leaves out my brand, what should I check first? Start with access. Check that your robots.txt allows retrieval bots such as OAI-SearchBot, PerplexityBot and Claude-SearchBot, and that your firewall or CDN doesn't block them. Then confirm that the facts you care about, like pricing or services, appear in the HTML before JavaScript runs. Fixing these removes blocks. It doesn't guarantee a mention, because engines decide what to include.