Amazon's new 75-character title limit: what it means for brands

On 10 June 2026, Amazon told sellers that from 27 July, product titles in every category except media must fit inside 75 characters, spaces included. Alongside the cap, Amazon added Item Highlights, a second field worth up to 125 characters, searchable and shown beneath the title in search results and on the product page.

The total indexable text barely moves: roughly 200 characters before and after. What changes is the shape. A single long string is now two fields with two jobs, and the way Amazon's AI reads a listing is why that matters more than the character count.

What actually changes on 27 July

From 27 July, any non-media title over 75 characters is out of policy. Sellers can shorten now or wait; listings stay active throughout.

Titles still over the limit after the deadline get rewritten to Amazon's AI recommendation, applied gradually. Brand-registered owners get a 14-day window in the Review Listings Changes panel to review, edit, or approve before anything goes live. Amazon's announcement does not confirm the same preview for sellers without Brand Registry, so don't assume one.

Item Highlights is new infrastructure. Amazon has confirmed it is searchable and surfaces next to the title. How heavily it is weighted for ranking, and how Amazon's AI surfaces draw on it, is undocumented and will only become clear through deployment.

The mechanical fit, trimming to 75 and parking the overflow at 125, is something Amazon's AI will do for free. What it won't do is decide which keywords were earning their position.

The mobile reason is true but incomplete

Amazon's stated reason is mobile: 75 characters keeps the full title visible on a phone. That's fair, and not new. Mobile search results have shown only around the first 75 characters for years, so much of any 200-character title stopped being seen on a phone long ago.

But mobile doesn't explain the second field. To clean up phone display, Amazon could have truncated the title and left it alone. Instead it split the title into an identity line and a separate, labelled, searchable attribute field. That shape, short identity in one place and structured attributes in another, is what Amazon's newer AI readers want. The cap reads less like a formatting tidy-up and more like Amazon reorganising its catalogue into fields its own models can trust.

The reader that changes the maths: Rufus, now Alexa for Shopping

A long, keyword-dense title is harder for a language model to interpret than a short, clean one.

Take a real title structure: Panasonic Nanoe Salon Hair Dryer with Oscillating QuickDry Nozzle, Diffuser and Concentrator Attachments, 3 Heat and 2 Speed Settings for Styling, Japanese Technology for All Hair Types, EH-ANA6HN (Black/Pink Gold). For keyword search, that string did useful work. For a model establishing what the product fundamentally is, it's a lot of competing signal. A title built as [Brand] + [product type] + [single key differentiator] gives the model a reliable anchor; a keyword pile gives it ambiguity.

This matters because of how Amazon's AI shopping surface works. Rufus became Alexa for Shopping on 13 May 2026, now answering shoppers directly across the search bar, app, and website rather than from a separate tab. It runs on retrieval-augmented generation: before it answers, it pulls current information from the catalogue, reviews, and Q&A, and grounds its response in what it retrieves. Listing quality therefore decides whether the assistant recommends your product or a competitor's.

For an intent-led query like "best hairdryer for curly hair," the assistant matches products against a need state. A clear identity line lets it classify, compare, and cite the product with confidence. The assistant is conservative by design and avoids recommending products it cannot confidently explain, so a title it can parse cleanly lowers the risk it skips you.

Where displaced keywords and claims should go

If a title sheds value propositions to hit 75 characters, those propositions need a home where the AI still reads them as attributes, not noise.

Item Highlights is the first home for displaced terms. Amazon has confirmed the field is indexed, which puts it in the retrieval layer, and its 125-character limit nudges towards the compact, attribute-style phrasing models parse well. Lead with what Amazon asked the field to carry, materials and recommended use cases, then work displaced keywords into that phrasing rather than dumping them in as a second keyword pile. Use-case and audience language does double duty, because that's what the assistant reaches for when answering who a product suits and when they'd use it.

Two constraints: Amazon's guidance is to use benefit-driven phrases rather than full sentences and not repeat the title, and Item Highlights only display once the item name is under 75 characters. Trimming the title is the precondition for the field paying out at all.

There's also a quieter destination. Benefit-led claims and use cases may sit better in structured backend attributes than in any front-facing field, because that's where Amazon's COSMO knowledge graph reads from when it maps a query to a product. COSMO works against fields like item type, intended use, material, and compatibility, and incomplete backend attributes are a real liability there. Whether a given claim belongs in Item Highlights, the bullets, or backend categorisation is something we're still researching at Azoma, and we'd rather flag that than pretend the priority order is settled. What's clear is that "put it in the title" is no longer the default answer.

Keyword density was a search-engine habit. The new game is structured attributes

The deeper shift is away from the keyword string as the unit of optimisation. To answer "is this hairdryer compact" or "can I take it travelling," a model doesn't want a denser title. It wants structured, consistent, attribute-rich data across fields it can parse, compare, and cite. Keyword density in a title is a legacy search-engine concept. The likely winner under the new structure is well-populated backend categorisation, the kind held in the VSSC file, carrying more of the weight long titles used to carry alone.

That reframes what to prioritise. The most query-adjacent, structured fields, title plus Item Highlights plus core PDP copy, are likely to become the primary retrieval stack for AI answers. Bullets, reviews, and Q&A stay important as the substantiation layer the assistant draws on for longer, reasoned answers.

Does Item Highlights move consideration?

On AI-mediated consideration, plausibly yes. Because the field is indexed and feeds the retrieval layer, it can influence whether the assistant mentions your product at all when responding to a prompt. That matters, given that shoppers who engage with the assistant convert at materially higher rates. We're speaking only to AI search visibility here, not organic click-through.

What to do before 27 July

Most sellers are looking at dozens or hundreds of non-compliant listings with about six weeks left. Spend manual effort where rank and AI visibility live, and let low-stakes listings ride.

1. Inventory the problem. Export your catalogue as a flat file from Manage All Inventory and flag every title over 75 characters. Prioritise by traffic and by whether rank depends on words past character 75.
   
   

2. Win back the easy characters first. Use numerals, abbreviate units, cut filler. Most titles surrender ten to fifteen characters this way without losing a real keyword.
   
   

3. Treat the 75 as an identity decision. Lead with brand, core product noun, and the single differentiator that earns the click. Everything else is a candidate for Item Highlights, bullets, or backend attributes. A structure like Panasonic Ionic Hair Dryer 1875W with Diffuser reads well on a phone and parses cleanly for a model.
   
   

4. Build Item Highlights deliberately. Lead with materials and use cases, weave displaced keywords in as natural attribute phrasing, don't repeat the title, and remember the field only shows once the title is under 75 characters.
   
   

5. Strengthen the backend. Keep item type, intended use, material, and compatibility fields complete and consistent. This is where COSMO reads, and where much of the old title's ranking work is likely to migrate.
   
   

6. Brand owners, guard the window. Put the 14-day Review Listings Changes review on the calendar so no automated cut goes live unwatched. Where a listing barely sells, letting the AI recommendation stand costs little.
   
   

Summing Up

After 27 July, every over-length title gets rewritten. The only open question is who does it. Amazon's AI optimises for length and basic relevance; it doesn't weigh which keywords carry your rank or run the result through a compliance check. Engineer the 75/125 split yourself and those decisions stay with you. Leave it to the default and you find out, gradually, what the algorithm decided your listing was about.

This is what Azoma does at scale: structuring titles, Item Highlights, and backend attributes across a full catalogue so every surface reading your listing, the keyword engine and the AI assistant alike, gets a clean, consistent signal. With hundreds of non-compliant titles and six weeks to go, that's a catalogue project, not a copy task, and one worth getting right before the default runs on your behalf.