In this Post:
- Why AI Product Recommendation (AIPR) Matters
- How AIPR Works (Short Version)
- Visual Explanation (AIPR Atlas)
Why AI Product Recommendations (AIPR) Matters
People are increasingly using AI to ask questions.
This AI adoption is changing how people find answers and solutions.
This is a big market landscape change for all industries.
AI changes how products are discovered – in a way that fits humans’ most natural behaviors to ask questions.
It presents the need for companies to invest in a long-term strategy.
First to understand how AI perceives question narratives, then understand how AI recommends solutions and products that fit the narrative.
In this process, AI exhibits a pattern of behavior to “find the right answer”
How does a product fit into a solution narrative?
How can you see evidence in real life that prove this?
How can you make this easy for AI to discover this? Remember this?
This is Narr Theory’s AIPR Strategy in the broader domain in the Voice of AI.
The Key Behavioral Shift – Narratives
The main difference between AI and Google search – is the user behavior.
On Google, people enter keywords.
In AI, we have conversations.
Instead of putting in keywords, we can ask “what product should I buy?” or “what solution is best?”
People can ask for comparisons, follow up questions, technical details, and even social validations.
We can leave the conversation for a moment and come back to it later.
We can even use voice and “talk to AI” in the process.
This stresses the importance of how solutions appear in narratives.
Here is an example.
Keyword Search:
“best car for day-to-day”
Narrative Search:
“I need to buy a new car. I will be using it for day-to-day commute so I want it to be fuel-efficient and easy to maintain. I want a car that won’t break down easily. Preferably a red car but white and silver are also ok. I prefer SUVs but sedans are ok if it fits my budget better. I am looking for values around $55-80K – what do you recommend?”
Notice…
- There are technical constraints (fuel efficiency, maintenance, color, price)
- Tradeoffs (fits my budget better)
- The Narrative (how it all fits into a conversation)
So what happens when AI is in the position to recommend a product?
AI Product Recommendation (AIPR) Explained
Understanding how AI recommends products is similar to how humans recommend – they pass on the narrative of what they know.
Since AI does not have hands-on experience with products like humans do, AI reads “what others say” at supercomputer scale and synthesize the “Best Recommendation” from the corroboration of evidences.
How AIPR Works (Short Version)
Below is the high level representation of the step-by-step behavior that AI systems tend to exhibit when recommending a product. This is not, however, the “exactly how every single AI system thinks” or “every how AI behaves exactly, 100% of the time.”
While Narr Theory strives to obtain such accuracy in understanding AI’s thinking and behavior patterns, similar to how AI can tell if a product is the “best in this narrative”, we strive to get closer and closer to the 99% confidence in the overall approach to how this can be examined, understood, and proven.
The below framework is derived from repeated observations, testing, and carefully curated corroboration with first-party authority explanations where available (e.g. OpenAI), rather than having direct access to proprietary secrets that only the publishers of truth like OpenAI or Alphabet can access.
If you want a more detailed breakdown, visit Narr Theory’s AIPR Atlas here.
- User Asks a Question (What product do you recommend for XYZ?)
- This creates a narrative by the user including things like:
- What they need
- Why they need it
- Where
- How
- Constraints, deciding factors
- Tradeoffs
- This creates a narrative by the user including things like:
- AI “Understands” the Question
- AI remembers contextual information of the user (if available)
- Their occupation
- Their experience with XYZ
- Their preferences
- AI also puts it into context of the “real human world”
- The corresponding answer should resemble practical human application and limitations.
- E.g. “something about size..? like drill for food truck”
- AI remembers contextual information of the user (if available)
- AI Identifies and Breaks Down Constraints & Tradeoffs
- AI is great at putting user’s questions into real human context.
- A user asks “car for commute”
- AI may break it down to “fuel efficient”, “easy to maintain”
- “easy to maintain” – parts network
- AI may break it down to “fuel efficient”, “easy to maintain”
- User mentions “best for my budget”
- ex. Tradeoff between fuel efficiency vs. price
- A user asks “car for commute”
- AI has built in heuristics that model the human world. Many times, AI may suggest “best value for budget” options, or the extreme-end option like “the most fuel efficient car”.
- Often, AI will consider factors not mentioned specifically in the prompt
- If the user has accessibility issues, AI may recommend products that fit that criteria.
- Often, AI will consider factors not mentioned specifically in the prompt
- AI is great at putting user’s questions into real human context.
- AI Looks Within Its Internal Model Memory First
- Every time a prompt is entered, AI is using its latest training module to action on the prompt.
- AI in fact have some core internal memory for patterns of questions like “how many fingers do humans have?” and even some brand-related heuristics like “what is a good shoe brand?” (Nike, Adidas)
- AI will first see if they know the answer to the question from “what I already know”.
- If the “knowledge” is not within this internal model memory, AI performs a web search.
- The more long-tail or niche a narrative, the more likely the “knowledge” will need to searched.
- AI in fact have some core internal memory for patterns of questions like “how many fingers do humans have?” and even some brand-related heuristics like “what is a good shoe brand?” (Nike, Adidas)
- Every time a prompt is entered, AI is using its latest training module to action on the prompt.
- External Knowledge Search – Web Search or AI Search
- Some tools like ChatGPT or Google AI have their own integrated web search service.
- AI will turn your “decision criteria” or “search narrative” into a different format that will give them the best chance of retrieving the best results from their web search. (AI Web Search Point)
- Sometimes, AI may perform this multiple times simultaneously depending on what system you use and what membership you have (e.g. ChatGPT Premium), and the level of truth-confidence that AI targets to achieve (80% vs. 95% confidence).
- While it is uncertain to know exactly how web page results show up on AI’s web search, we can presume that it resembles how traditional search engines work.
- Sometimes, AI may perform this multiple times simultaneously depending on what system you use and what membership you have (e.g. ChatGPT Premium), and the level of truth-confidence that AI targets to achieve (80% vs. 95% confidence).
- AI Corroboration Synthesis
- This is particularly important for the non-technical descriptions of products – the narrative.
- “This product was… most useful, practical, beginner-friendly, easy to use, visually appealing, children loved it, popular, has good reputation, recommended by, best performance… etc.”
- This is the part that especially set AI apart from humans.
- AI will scan through multiple web pages and look for clues that corroborate or contradicts a certain claim.
- AI will then synthesize its findings to form its recommendation(s)
- This is particularly important for the non-technical descriptions of products – the narrative.
- User Feedback, Recalibration, and Ongoing Conversation
- The special thing about AI Product Recommendation (AIPR) is not just that it can find information at superscale. It’s also its conversational relationship with humans.
- AI can incorporate the user’s personal context from previous conversations into the decision narrative, although not explicitly mentioned.
- AI is a great solution for people to discuss their workflow and leisure activities. People have conversations with AI on “how to do” conversations, “how to be better at”, “this problem to solve”, and other types of conversations that may have to do with a certain topic.
- This ongoing conversational nature presents more than one opportunity for the user to consider a product purchase, while their decision criteria potentially gets narrower and more specific.
- AI Product Recommendation
- Even after a long conversation with the user, the decision-conversation reaches a diminishing point, where any new input by the user, unless the new information presents a new variable, has diminishing impact on the decision criteria.
- At this point, AI’s position on its recommendation may not change much.
- AI Product Recommendation is then ultimately formed from either the initial, short-conversation exchange or from a much longer and personalized conversation with the user.
- Even after a long conversation with the user, the decision-conversation reaches a diminishing point, where any new input by the user, unless the new information presents a new variable, has diminishing impact on the decision criteria.
AIPR Atlas
Narr Theory’s AIPR Atlas is an interactive visualization of how AI recommends a product from the user question, AI question interpretation, narrative breakdown, knowledge retrieval, corroboration synthesis, and to the final AI product recommendation.
This can help visualize what gaps may currently be existing for your product’s likelihood of being recommended by AI, and why competitors are being recommended instead.
Visit the Atlas here.
Key Takeaway
Whether a user discuss a product recommendation with AI for 10 seconds, 10 minutes, or 10 months – the main advantage point for AI is personalization.
As humans’ adoption of AI increases, AI’s overall usefulness for real-human conversation and everyday application is also likely to improve.
This presents moments in conversations between humans and AI, where AI is in the position as the influencer of knowledge, perception, and decisions.
Having a product recommended by AI is not as simple as optimizing your website’s GEO (this has minimal impact, in fact, as it only controls the websites you own) or paying web pages to talk about your positively.
It is important to understand and leverage both the tactics that will help with AIPR in short term, and invest in acquiring knowledge, skills, and experience narrative intelligence tools that can help you understand your products’ and brands’ position in the every-day narratives, especially at the points of authority and influence, long term.
AI may show exponential improvements in the next 3-5 years in its ability to find more accurate, truthful, and credible information for questions.
For this reason, Narr Theory hypothesizes that companies that leverage AI’s ability to show how people are asking questions and searching for solutions will have be advantageous over others.
It may become essential to develop initiatives to keep a pulse on strategic analysis and discussions around:
- AIPR Gap Analysis
- Narrative Product Position Strategy
- Unfulfilled Narrative Whitespace
- Voice of Customer (Narrative Focus)
- Voice of AI
Narr Theory, the publisher of this post (not AI generated) and the creator of the AIPR Atlas, specializes in helping companies understand their product and brand’s position in AIPR.
Schedule an introductory call to see where you can start.
Into the Future
In the long-term, it is important to be aware of the the motif of AI developers to help predict how the AIPR methodology may evolve in the future. At the rate of technology advancement, “tomorrow” may come sooner than later.
Around June 2026, ChatGPT officially released its first advertisements on the platform. This signals OpenAI taking action on its motif of monetization, and willing to leverage its position in the market.
Narr Theory’s AIPR Participant Motif explores these ideas that may influence the landscape both in immediate and distant future.
AI systems are intelligent, abstract, and dynamic in nature while the creators of AI systems have the motif not to reveal their key secrets to the public.
Narr Theory is dedicated to researching the latest and most proven methodologies in AIPR that is congruent with the first-party publishers of truth (OpenAI, Google) at the highest level of accuracy, and other sources of repeatable and reliable sources of corroborated truth deriving from our epistemic core.
To stay posted on our thought leadership, follow us on LinkedIn or subscribe to our mail list here.
Leave a Reply