On 27 July 2026, Amazon did something it had never done before. It started rewriting brands' product titles with its own AI. Titles in every category except media were capped at 75 characters, down from 250. A new field called item highlights appeared, worth another 125 characters. Sellers got a 14-day window to approve Amazon's suggested changes, after which the platform applied them anyway (Amazon Seller Central).
Ten days after the deadline, the picture was messy. On a recent MerchantSpring Marketplace Masters episode, Alasdair McLean-Foreman, founder and CEO of Teikametrics and one of Amazon's earliest third-party sellers back in 2002, described it plainly: a mixed bag, and concerning for a lot of sellers. Some reported sales drops in the Seller Central forums. Some watched Amazon's algorithm strip their brand name out of the title entirely. The early read is that outcomes depend heavily on how good your listing was going in.
The character count is the visible part. The argument worth your attention is what it signals underneath, because the title cap is a forcing function for a much larger shift in how Amazon expects products to be described.
For years, the 250-character title was prime real estate, and an entire cottage industry grew up around optimising it. Amazon has now split that space in two: up to 75 characters for the title, and up to 125 for item highlights, a separate field for attributes and use cases. Roughly the same total space, carved into two jobs.
The pain came from Amazon doing the carving. Its AI decided what survived the cut and what moved where, and it did so inconsistently across listings. Sellers who had packed keywords, sizes and specs into a long title found the machine making choices they would not have made. That is the real lesson of the past ten days: Amazon will automate this, it will apply a fairly cookie-cutter algorithm across the platform, and the brands that had thin or disorganised content had the least control over the result.
McLean-Foreman's thesis is that the title rule is a top-down move toward agentic commerce. Rufus, Amazon's shopping assistant, has been folded into Alexa for Shopping, and the title deadline followed weeks later. The two are connected. To answer full, conversational questions instead of two-word searches, Amazon's AI needs better structured, machine-readable data about what each product actually is and does. Shortening titles and creating a dedicated attributes field is how it gets that data at scale.
The precedent is Google. As shoppers move from typing keywords to asking questions, AI-generated answers sit above the old blue links, and every AI-mode answer is one that a shopper did not have to click on a paid result to get. Amazon is building its own version of that with Alexa for Shopping, and Teikametrics says it is already seeing more of its customers' search-to-listing traffic arriving through it. This is not the usual Amazon habit of shifting the goalposts. It is deliberate infrastructure for conversational shopping.
The most useful line from the conversation was also the bluntest: if six hours of battery life is not in your attributes, no AI rewrite will put it there. Amazon's algorithm can only reorganise what you have given it. If your listings are thin, reorganising thin content produces thin results.
Consider how search itself is changing. “Best wireless earbuds” is three broad words where the shopper does the filtering. The conversational version is “wireless earbuds that stay in during running with at least six hours of battery.” That specificity, the use case and the attributes behind it, is exactly what belongs in the 125-character item highlights field, and exactly what Alexa for Shopping indexes to match a product to a question. If it is missing from your data, you are not in the answer.
The framework is separation, not repetition. Use the title to differentiate the product name and brand. Use item highlights to differentiate use cases and key attributes. Most brands already hold this information, but they have historically buried it in the product description, often pulled from a PIM or central catalogue system. The move now is to lift core selling characteristics and use cases out of the description and into item highlights, where the AI will actually read them.
Practically: identify the two or three attributes that genuinely drive purchase in your category, write them as use cases rather than buzzwords, and keep them out of the title so the title can do its own job. Repeating the same phrase in both fields wastes the space Amazon just handed you.
Here is the part most teams miss. Your advertising data already tells you which words drive demand. The keywords you bid on are demand signals, and McLean-Foreman argues the sophisticated operators are feeding those signals, along with search-query reports and market-share data from tools like Jungle Scout, Helium 10 and Stackline, into AI to rewrite listing content. Better ads inform better listings, which lift organic and Alexa visibility, which improves return on ad spend. A flywheel, not two separate workstreams.
That has an uncomfortable implication for how many brands and agencies are organised. Content and advertising usually sit in different teams, sometimes different P&Ls. When content moves to the front of the funnel, that separation leaks value. The advice is direct: break down the silos, and put the teams that are already fluent in keyword optimisation to work on content.
The biggest gains show up in the long tail. Most operators live by the 80-20 rule and pour their effort into the top 20% of listings. But a brand with, say, a thousand SKUs that has only worked the head has a large dormant tail. Re-indexing those listings can reactivate SKUs that were effectively invisible. Teikametrics reports lifts as high as 135% in that scenario, because Amazon rewards fresh indexing. Read that as a first-party result, not a guarantee: the company was careful to say there is no single number that applies to everyone, and a seller with five already well-optimised products should expect far less. The upside scales with how much neglected catalogue you have.
Walmart is the closest parallel. Its marketplace architecture mirrors Amazon's best practices, it has Sparky as its own shopping assistant, and it is growing quickly. Brands that tried Walmart two or three years ago and found it underwhelming should look again, because the return for getting ahead of the curve there is currently high, and the operational playbook is familiar.
TikTok Shop is a different game. It is a discovery platform driven by creators and content, not listing optimisation, so the lever is finding the right creators and producing the right video rather than tuning attribute fields. And ChatGPT sits on the horizon as another surface where product data will need to be legible to AI. The common thread across all of them is structured, machine-readable content. McLean-Foreman also expects AI-generated imagery to matter soon, since the economics of updating an image with AI now undercut a full photo shoot.
Start by checking what Amazon did to your titles during the rewrite, especially whether your brand name or any load-bearing attribute got dropped. Then treat item highlights as real estate, not an afterthought: move genuine use cases and differentiating attributes into that field across your whole catalogue, not just the hero SKUs. Pull your advertising search-query data into the content process so the words you already pay for shape the listings. And prioritise the long tail, where the neglected upside sits.
Amazon's advertising business is approaching 80 billion dollars in annual revenue, and it mandated this change, knowing Alexa for Shopping may cannibalise some of its own sponsored placements, the same trade Google accepted with AI mode. A company does not do that casually. Expect more consumer search to move into conversational shopping through the back half of the year, and expect this Q4 to send more sales through Alexa, Sparky and ChatGPT than any before it. The brands that get their content ready now will be the ones the assistants can actually recommend.
The July rewrite runs on the same fundamentals you can already measure: listing quality, velocity, reviews and rank across every account. The brands that come out ahead are the ones watching those signals, not guessing which SKUs Amazon quietly reshaped. MerchantSpring gives sellers, agencies and enterprise brands one view of marketplace performance, so you can spot the listings dragging your numbers down before they cost you a sale. See it on your own accounts with a walkthrough of your data.