Allow me to explain how e-commerce brands can improve their visibility in AI Search answers by using agentic commerce.

In the past, with lexical search, when we were searching for a product and it was available or not. That was the end of the buying story. It was binary. Today, AI shopping assistants can suggest a better product than the one you were searching for because they “understand” your needs.

The AI can suggest a new item, a new brand. Product discovery changed thanks to semantic search. Semantic search is the future of marketing. But until now, “AI Search” still required you to click a link to leave the platform and visit a website.

All e-commerce journeys looked the same, with search engines and LLMs: 

→ Search 

→ Get answer 

→ See links (hopefully)

→ Leave platform (Google Search, ChatGPT, etc)

→ Add to cart

→ Buy

Agentic commerce can change this.

Traditionally, the marketing funnel involved 4 layers:

1️⃣ Awareness

2️⃣ Interest

3️⃣ Desire

4️⃣ Action

I’ll explain how we can compress the four stages of the buyer journey into a single natural conversation, without leaving the AI platform.

But first, I’ll show you how with the Agent2Agent protocol (and AP2), users can discover and buy directly within Perplexity using their PayPal account. 

❌ No leaving the platform

❌ No new logins

❌ No friction.

I’ll show you how this gives us a competitive advantage in AI visibility.

Of course, the four stages of SEO all remain:

✅ Discovery

✅ Crawling

✅ Indexing

✅ Ranking

We’re not replacing this, we’re adding a faster alternative!

Instead of AI crawlers processing websites to guess whether an item is available, I’ll show you how Perplexity sends a structured request to PayPal’s index, which returns our inventory and exact pricing instantly.

I’ll explain how we grounded the AI models.

PayPal creates what I call a “state index,” a transactional database that represents the state of our catalog (the physical reality of our warehouses).

Perplexity then ingests our product feed to build its vector index. Perplexity converts our product descriptions into mathematical representations (embeddings) so the AI can understand the “intent” behind a product, not just keywords. That is what we call semantic search.

Here are the takeaways of my video:

1. There are actual ways to improve the AI visibility of brands (beyond server-side rendering).

2. Our team built APIs specifically for PayPal

3. Remember that Perplexity is an AI platform, not a language model: you can switch between LLMs using Perplexity’s settings

5. The challenge was to ground whatever AI model is used in Perplexity.

6. To achieve this, we sent a modified version of our product feed to PayPal which then used our dataset to ground the LLM of your choice in Perplexity

7. PayPal built a state index which in some way looks a little like the Google Search index

8. Perplexity then built its own vector index, which is a semantic graph (a vector database) 

9. Thanks to the Agent2agent protocol, both PayPal and Perplexity interact

10. To simplify, think of two assistants calling each other back and forth 

11. Perplexity is the buying agent, while PayPal is the selling agent  

12. Using a daily updated version of our product feed, we are grounding whatever large language model you decide to use in Perplexity 

The result? Improved AI visibility: the Adorama brand stands out in the AI answers and shopping results, as I’ll demonstrate in this video!

With this approach, we bypassed all the technical layers associated with crawling and indexing.

We plan to systematise this approach to other AI shopping solutions. We won’t stop here and we’ll keep improving because we believe agentic commerce is part of the future, not as a replacement but as an addition.

We believe in offering users the choice. We want them to choose whatever search experience they prefer. This is just the beginning. We’ll continue to innovate and create, no matter what!

Ask me your questions in the LinkedIn post and remember the benefits of using product feeds to ground models is an improved AI visibility and better Share of Voice.

It was the hard work and brilliant contributions from our fantastic Adorama teams, guided by our CTO Hani Batla, that brought this to life! 🙂