Amazon shipped Rufus to every Indian buyer last year. Flipkart followed. Myntra and Nykaa are testing their own. These marketplace AI agents are not chatbots bolted onto a help page. They sit inside the buyer flow, between the search bar and your listing, and they decide what the buyer sees about your product. Most brands have not noticed. Their listings still read like it is 2022.

What the agent is doing

A buyer types "face wash for oily skin under 500" into Amazon. Before the agent existed, search ran keyword matching on titles and showed a ranked list. The buyer scrolled, read bullets, made a call. 

Now the agent runs first. It reads the query, scans the top forty results, pulls bullets and reviews from each, and writes a short answer at the top of the page. "Three options stand out. X is gentle and budget friendly. Y has stronger salicylic acid for active acne. Z is unscented." 

The buyer reads that paragraph. Sixty percent of the decision is already made. 

Your listing is no longer competing for a click. It is competing to be the listing the agent quotes. 

What the agent reads

The agent does not read your title the way a human does. It reads it as data. Title, bullets, A+ content, backend keywords, structured attributes, review text, return rate, question and answer section, price stability, stock history. 

It scores you on three things. 

Coverage. Does the listing answer the questions buyers in this category ask. For face wash that means skin type, key actives, fragrance, pH, dermat tested status, pregnancy safety, shelf life. If your bullets say "premium ingredients for healthy skin" the agent has nothing to work with and skips you. 

Trust. Return rate, negative review themes, seller rating, buy box history. A listing with a 14 percent return rate gets quoted less even if the bullets are perfect. 

Price clarity. Erratic pricing, missing MRP, unclear pack size. The agent treats this as a signal of low quality and pushes the listing down. 

What changes for the brand

Three things break under the new system. 

Keyword stuffed titles hurt you now. "Best Face Wash for Oily Skin Acne Pimple Control Anti Aging Glow Skin Whitening" used to win search. Today the agent reads it, flags it as low signal, and ranks the cleaner title above you. Amazon listing optimization in India is no longer a keyword density exercise. It is a clarity exercise. 

Bullets written for humans skimming on mobile are not enough. They have to answer the question the agent is going to be asked. If a buyer asks Rufus "is this safe during pregnancy" and your listing does not address it, Rufus pulls the answer from a competitor that does. You lose the sale to a brand whose listing was written for the agent. 

Ad spend now competes with the agent's recommendation paragraph. You can win the sponsored slot and still lose the buyer because the agent suggested another product two inches above your ad.   

What working with the agent looks like in practice

A skincare brand on Amazon. Twenty crore annual run rate, flat for six months, ad ROAS dropping every quarter. 

Audit pulled the top fifty SKUs. Forty one had bullets written before Rufus existed. None of them answered the eight questions buyers ask in that category. Reviews mentioned fragrance complaints that the listings did not pre empt. Three SKUs had buy box loss above 30 percent because of erratic third party seller pricing. 

The fix took six weeks. Bullets rewritten to answer the eight questions in plain language. A+ content restructured around skin type rather than brand story. Backend keywords cleaned. Buy box defended through seller controls. Negative review themes addressed in the listing copy itself, so the agent stops surfacing the objection. 

Organic units up 34 percent on the same fifty SKUs. Ad ROAS up 22 percent because the listing now closed the buyer the ads brought in. 

Nothing about this is exotic. It is listing work, catalog work, review work. The difference is who the listing is being written for. 

The questions to ask this week

Open Rufus. Search for your top SKU the way a buyer in your category would. Read the paragraph the agent writes. If the answer is wrong, thin, or quotes a competitor, the listing is the problem. Not the ads. 

Pull your top ten SKUs. List the eight questions buyers ask in your category. Count how many your listing answers. Most brands answer three. 

Check return rate by SKU. Anything above the category average is dragging your rank in ways the agent compounds every week. 

This is not a strategy exercise. It is a Tuesday afternoon. 

The shift

Marketplaces used to reward the brand that bid hardest and stocked deepest. They still reward that. But there is a new gatekeeper between the buyer and the listing, and it reads everything you have written. Any serious D2C marketplace strategy from here has to start with what the agent sees first. 

The brands that win the next two years on Amazon and Flipkart are the ones rewriting their catalogs for the agent now, while most of the category is still optimizing for a buyer who no longer reads the page first. 

If any of this sounds like your Amazon dashboard right now, let's talk. Thirty minutes, no slides. We'll tell you whether it's worth fixing or whether ads will get you there faster.