Amazon cut its title limit to 75 characters, retired Rufus in favor of a unified Alexa for Shopping assistant, and started rewriting non-compliant listings with its own AI model, sometimes without asking. None of this is on the horizon. It already happened. Here's what actually changed, and how to write for the system that replaced the one you learned on.
Three things, all within the last three months, and all pointing the same direction: away from keyword matching, toward intent understanding.
On June 10, 2026, Amazon posted a Seller Central policy update titled "Updates to improve your product titles begin on July 27." Effective that date, titles in every category except media (books, music, video) are capped at 75 characters, including spaces and the brand name, less than half the roughly 200-character ceiling sellers had worked with for years. Alongside the cap, Amazon introduced a new Item Highlights field: 125 additional searchable characters displayed below the title, keeping total indexable space at 200 characters but splitting it into two purpose-built fields instead of one long, keyword-stacked string.
Titles that exceed the new limit don't just get flagged. They get rewritten, by an Amazon AI model that pulls from the existing listing and Amazon's own relevance signals, on Amazon's schedule, not yours. Brand Registry owners get a 14-day window to review, modify, or approve the recommendation before it goes live. Sellers without Brand Registry don't get that window; the change applies directly.
At the same time, the assistant reading and recommending that content changed too. Amazon retired Rufus, its original AI shopping assistant, on May 13, 2026, folding it, together with the Alexa+ assistant, into a single, unified experience called Alexa for Shopping. It's now the default AI assistant for every signed-in U.S. customer on the Amazon Shopping app and website, no Prime membership or Echo device required, and it can compare products, track prices, and act on a shopper's behalf rather than just answering questions about them.
The title cap dropped from about 200 characters to a hard 75. Total indexable space stays at 200, rebalanced into a short title plus a new 125-character Item Highlights field.
When Amazon's AI rewrites a non-compliant title, Brand Registry owners get 14 days to review or change it first. Sellers without Brand Registry get no window at all.
Because the system underneath both search and Alexa for Shopping isn't matching your keywords anymore. It's trying to understand why someone is searching in the first place.
That system has a name: COSMO, a large-scale commonsense knowledge graph Amazon's own research team published in 2024 and has continued building on since. COSMO doesn't just index what a product is, it connects products to the real-world needs, activities, and situations they solve for, in structured relationships like "slip-resistant shoes, used for, pregnant women." At last published scale it spans roughly 6.3 million nodes and 29 million of these knowledge connections across about 18 product categories. That knowledge graph is what both Amazon's own search ranking and Alexa for Shopping draw on to decide whether your listing actually answers what a shopper meant, not just what they typed.
The practical result: Amazon's A9/A10 search algorithm and Alexa for Shopping now run in parallel, and neither replaces the other, so you have to write for both at once. A title or bullet built as a string of keywords reads fine to the old system and poorly to the new one. Content that leads with who the product is for and what problem it solves reads well to both.
Amazon's knowledge graph connects a product to the situations it solves for, so a listing that answers a real need outranks one that just repeats a keyword.
Less compression, more clarity, and a different way of thinking about what each field is for.
| Content Element | The Old Playbook | The 2026 Playbook |
|---|---|---|
| Title length and style | About 200 characters, front-loaded with keyword variants, mechanical structure | 75-character hard cap; leads with use case and target customer; reads as a sentence a person would say |
| New structured field | Did not exist | 125-character Item Highlights field carrying supporting detail the title no longer has room for |
| Bullets | Feature lists, keyword repetition across bullets | Benefit-first, declarative statements that answer an implied shopper question |
| Backend search terms | Every keyword variant possible, including ones already in the title | Reserved for synonyms and spelling variants not already used naturally in visible copy: repeating a term already in the title wastes the space |
| Discovery and Q&A content | Optional, rarely prioritized | Discovery attributes (subject, audience, intended use) completed fully; Q&A section built out, since Alexa for Shopping reads and can answer directly from it |
What to check on your listings this week.
Do it now, so you aren't waiting for Amazon's AI to rewrite titles for you on its own schedule.
If you don't have it, you have no review window before an automated rewrite goes live.
Use the new 125-character field instead of trying to compress everything back into the title.
Remove keywords already present in your visible title or bullets, and use the freed space for genuine synonyms.
Answer the questions shoppers actually ask about the product, since Alexa for Shopping reads and can respond from it.
The growth curve on AI-assisted shopping suggests this isn't a niche channel to revisit later, it's becoming a primary discovery path.
Reported figures vary by source, but they point the same direction. Multiple industry analyses put 2026 U.S. voice commerce in the tens of billions of dollars, growing at a compound rate in the twenties percent annually. Roughly half of U.S. consumers say they've used voice search for shopping. Separate reporting on AI-referred retail traffic shows triple-digit year-over-year growth, converting meaningfully better than traditional search referrals. And in consumer surveys, a strong majority of shoppers say they've already used AI somewhere in a shopping journey, comparing prices, summarizing reviews, or getting product ideas, with more than half saying they've knowingly used a retailer's own AI assistant, Amazon's included.
None of these figures are certainties, and estimates differ meaningfully across sources, we're not treating any single number here as gospel. What they converge on is directional and hard to dismiss: the share of product discovery happening through conversational, AI-mediated interfaces is growing quickly, on a platform that just rebuilt its own content rules around exactly that shift.
Exact numbers vary widely by source, but they converge on the same shape: discovery through conversational, AI-mediated interfaces is climbing while traditional search stays flatter.
Most brands find out about a change like the 75-character cap after Amazon's AI has already rewritten their titles for them.
In-Syte, our proprietary intelligence layer, tracks structural changes like this across every account we manage, flagging non-compliant titles, incomplete discovery attributes, and thin Q&A sections before Amazon's own AI makes the call for you instead. That's the difference between adapting to a policy change on your terms and finding out about it in your performance dashboard three weeks later.
A Marketplace Assessment is a clear-eyed look at your current listings against Amazon's 2026 content standards, not a pitch.
Get Your Free Marketplace Assessment