Why does the same shopping query produce different AI recommendations?

AI responses can vary by session, platform and wording. One observation is a data point, not a verdict. Repeated, controlled testing across multiple sessions and platforms is the only reliable way to establish an honest baseline.

Independent Research

Insights on AI Product Discovery

Q: How to check whether AI assistants mention your products

Choose realistic buyer-style queries, run them across ChatGPT, Gemini and Perplexity, and record the platform, date, exact query wording and what was said about your products versus competitors. Document conditions clearly enough that someone else could repeat the test.

Explore how AI-powered buying journeys influence product discovery, brand representation and competitive visibility. Evidence-led perspectives for Indian D2C brands.

This is independent research, not client case studies. We run our own queries, note the date and conditions, and flag where results are uncertain or likely to change.

Published Insights

Research and Practical Guides

Evidence-based articles on how AI assistants handle product recommendations for Indian D2C brands, how to test your own AI product discovery visibility, and what the limits of that evidence actually are.

AI Product Discovery

Person testing AI assistant responses on a laptop in a Mumbai workspace

Practical Guide Independent Research

ChatGPT Gemini Perplexity

How to Check Whether AI Assistants Mention Your Products

Most brands have never typed a real buyer question into ChatGPT, Gemini or Perplexity to see what comes back. This article walks through a simple, repeatable way to test it yourself: choosing realistic queries, running them across platforms, and recording what is actually said about your products versus competitors.

It is a practical starting point, not a guarantee of what you will find. AI responses can vary by session, platform and wording, so we show how to document conditions clearly enough that someone else could repeat the test.

A step-by-step method for writing buyer-style queries, not brand-style ones
How to record platform, date, and exact query wording for reproducibility
What counts as a mention, a recommendation, and an omission
Common mistakes that make a test look worse or better than reality

Discuss This With Us

Product Representation and Accuracy

Comparison Independent Research

AI Product Discovery vs Traditional SEO: What Brands Can Actually Measure

Search engine optimisation and AI product discovery are related but not the same discipline. SEO measures rankings and organic traffic against known algorithms. AI discovery measures what a conversational assistant says, to whom, and why, under conditions that are far less transparent and more variable.

This article separates what is measurable today from what is still speculation, so marketing and growth leaders can set realistic expectations before investing time or budget.

Where AI product discovery overlaps with SEO and where it diverges completely
What can be observed and documented versus what cannot yet be controlled
Why keyword-style thinking does not transfer directly to conversational queries

Dimension	Traditional SEO	AI Discovery
Input	Keywords, backlinks, page structure	Conversational buyer queries
Output	Search ranking position	Product mention in a generated answer
Measurable?	Yes, consistently	Yes, with conditions documented
Controllable?	Partially	Not directly
Variability	Low to moderate	High across sessions

Explore the Sprint

Competitor Visibility in AI Answers

Methodology Independent Research

Why the Same Shopping Query May Produce Different AI Recommendations

Ask an AI assistant the same product question twice, on different days or platforms, and the answer can change. This article explains the known reasons behind that variability, drawing on our own repeated testing rather than assumption.

Understanding this variability matters for any brand trying to interpret a single good or bad result. One observation is a data point, not a verdict, and this piece explains why repeated, controlled testing is the only reliable way to read it.

Factors that can shift AI responses between sessions and platforms
Why a single favorable or unfavorable mention should not be over-read
How controlled, repeated testing produces a more honest baseline
What brands can reasonably conclude from observed patterns over time

Platform

High

Session

Medium

Wording

High

Date

Moderate

Indicative relative variability observed in ACTRISE testing. Not a controlled benchmark.

Discuss Your Brand

Apply This to Your Brand

Explore the AI Product Discovery Sprint

If these questions are relevant to your brand, the AI Product Discovery Sprint applies this same testing discipline directly to your products, categories and named competitors. It is a bounded, evidence-led engagement, not an open-ended audit.

See the exact scope, deliverables and what it does not promise before you reach out.

View the Discovery Sprint

What the Sprint includes

A fixed-scope, 4 to 6 week engagement that moves from baseline to re-measurement with eight documented deliverables.

12 controlled buyer queries across ChatGPT, Gemini and Perplexity
Up to 36 baseline observations documented with conditions
Up to 3 competitor brands measured under identical conditions
One controlled re-measurement cycle to show what changed
Eight deliverables including an executive summary

Schedule a Consultation