The Amazon Listing in the AI-Reader Era:
A Field-by-Field Guide

The change that forced the question

Amazon's stated reason for the cap is mobile. The majority of Amazon shopping happens on phones, and a long title truncates on a small screen — the shopper sees the first 70 to 80 characters and the rest is hidden behind an ellipsis. A 75-character title fits. Amazon also frames the move as bringing its titles in line with the title length other online stores use. Read at face value, it is a readability change.

The more consequential part is the field that arrives alongside it. Item Highlights gives a listing an additional 125 characters, and Amazon is explicit about two things: the content is searchable, and it is visible below the title in search results and on the product detail page. Amazon's own guidance is equally explicit about the format — Item Highlights should be written as comma-separated phrases, not full sentences. The worked example Amazon gives turns a sprawling title fragment such as "Compact charger for MacBook Air, iPhone, Galaxy, iPad, cable not included" into a tight highlights string: "USB-C, PPS Support, Cable not included."

So the question a digital commerce lead asks first — does the bigger field mean we shift to longer, more keyword-rich bullets? — has the wrong shape. Item Highlights is not bullets, and it is not sentences. It is a new searchable surface with a different job. The title's job narrows to identity. Item Highlights inherits the keyword, use-case, and claim work that used to be crammed into the title. The rest of this guide is about getting each field into its right job before Amazon does it for you.

A listing is read by three audiences now

Every text field on an Amazon listing is now read by three audiences, each with a different rubric. This three-persona model is the spine of the entire Genrise cluster; the strategic case for it lives in the digital shelf optimization piece, and the cross-assistant landscape is mapped in the AI shopping assistants field guide. The short version is what matters here, because it is the reason the title change is an architecture change and not a cosmetic one.

The human shopper is still roughly 85% of digital shelf traffic. They scan a title and an image on a phone, read a bullet or two, and decide in seconds. They reward a clean, legible title and benefit-led copy.

The AI-assisted human — the shopper asking Amazon Rufus, Walmart Sparky, ChatGPT, or Perplexity what to buy — is a smaller but fast-growing share. Amazon reported more than 300 million customers used Rufus during 2025. This audience never sees most of the page directly; they read what the assistant cites back, and the assistant reads structured, searchable, machine-legible content first.

The autonomous agent is a sliver today and rising. It evaluates product data programmatically and selects, or rejects, without a human reading the page at all.

The fields that serve these audiences are not interchangeable. A human rewards a readable title; an AI assistant rewards a searchable, claim-dense highlights string; an agent rewards clean, consistent structured attributes. The July change is Amazon formalizing that split — pulling the human-facing identity into the title and giving the machine-facing, searchable content its own field.

The title: identity, not keyword real estate

The title's new job is identity. In 75 characters, it should let a shopper recognize, differentiate, and compare the product — and nothing more. Amazon's own guidance prescribes an information order, and following it is the cleanest way to fit identity into the cap:

So "Amazon Fresh Decaf Colombia Whole Bean Coffee, Medium Roast, 12 oz (Pack of 3)" reads in the right order; "Medium Roast Decaf Coffee 12 oz Pack of 3 Colombia Whole Bean Amazon Fresh" carries the same facts in the wrong one. The constraint forces a useful discipline: lead with what identifies the product, drop what merely describes it.

The supporting rules all point the same direction — toward a clean identifier and away from a keyword dump:

• No repetition. A word can appear at most twice; brand names are subject to the same two-instance limit. Prepositions, articles, and conjunctions are exempt. "Levi's Men's 501 Original Fit Jeans" is compliant; "Levi's Men's Jeans Men's 501 Original Fit Men's Denim Jeans" is not.
• No decorative symbols. A defined set of special characters is prohibited outright, and others are allowed only in genuine contexts — a model identifier, a measurement. Star bullets and "★BEST SELLER★" framing are out.
• No subjective or promotional language. "Hot Item," "Best Seller," "free shipping," "100% quality guaranteed," and restricted phrases like "FSA/HSA eligible" do not belong in a title.
• Numerals, not words; abbreviations for measurement. "2-Pack" not "Two-Pack"; "55-inch" or "55 in" not "Fifty-Five inch."
• No ALL CAPS. Capitalize the first letter of each word, except prepositions, conjunctions, and articles.

For parent and child variations, size and color belong in the child title only — the parent stays generic, because the detail page shows the parent title until a specific variation is selected.

The shift to internalize: the title is no longer where keyword coverage happens. It is where identity happens. The coverage moves one field down.

Item Highlights: the new indexing surface

This is the field that changes the work. Item Highlights is a 125-character field that is searchable and rendered below the title in search results and on the product detail page. Because it is searchable, it is where the keyword, use-case, and claim content that used to live in the title now does its indexing work. Because it is rendered to shoppers, it also has to read cleanly. It is the one field on the listing that serves the human shopper and the AI reader in the same 125 characters.

Amazon's format guidance is specific and worth following literally: comma-separated phrases, not full sentences. The field is a prioritized list, not prose. That priority is the whole exercise — 125 characters is not enough room for everything, so what earns a place is a decision, not a dump.

A useful way to think about what goes in:

• The highest-intent keyword that no longer fits in the identity-only title.
• The target-persona use case — the job-to-be-done a shopper or an assistant is matching against ("travel size," "for sensitive skin," "office and home").
• The one or two claims worth indexing — the substantiated, specific attributes an AI assistant can cite with confidence ("fragrance-free," "USB-C," "dermatologist-tested").

A worked example, in Amazon's own pattern. Take a charger whose old 200-character title tried to carry everything:

The title now identifies the product in well under 75 characters. The Item Highlights field carries the compatibility, the standard, and the caveat as searchable, citable phrases. Nothing of value was lost; the keyword and use-case work was relocated to the field built to index it. The discipline of choosing which claims earn the space is where the product claims and AI visibility piece goes deep, and the reason claim selection matters is covered in the Amazon Rufus deep-dive — the assistant rewards specific, grounded claims and ignores vague ones.

Bullets: the conviction layer

Once keyword and claim coverage moves into Item Highlights, the bullet points are freed to do the job they were always best at: convincing a human shopper. Bullets are the place to translate attributes into benefits — to answer "why this one" rather than "what is this." Amazon's guidance is five bullet points on a standard detail page (A+ Content pages can carry more), each a sentence fragment without end punctuation, between roughly 10 and 255 characters, with semicolons separating phrases and no competitor comparisons or guarantee language.

Two practical notes. Numbers one through nine are spelled out in bullets — the opposite of the title's numeral rule — except in names, model numbers, and measurements. And bullets are where you reinforce the most important information from the title and Item Highlights for the human reader, not where you re-stuff keywords.

The craft of writing bullets that perform for the human shopper and the AI reader at once is its own discipline, and it is covered in depth in the product descriptions guide. The point for this architecture is narrower: with Item Highlights carrying the searchable coverage, bullets stop being a keyword overflow and return to being the conviction layer.

Description and A+: the context layer

The description is the context layer — 100 to 200 words that introduce the product and give it a place in its category. Amazon allows basic HTML tags, expects complete sentences in sentence case, and prohibits a long list of things that do not belong: price or cost information, company or vendor details, website links, ASINs, and a set of restricted phrases such as "eco-friendly" and "anti-bacterial."

The important structural fact is the relationship between the description and A+ Content. When A+ Content is active for a product, it replaces the standard description on the detail page. For enterprise brands with A+ across the catalog, the plain-text description is doing less display work than it once did — but it still feeds systems that read the raw catalog rather than the rendered page, which is exactly the machine-reading audience this whole guide is about. As with bullets, the writing craft sits in the product descriptions guide; the role in the architecture is to give the product context that the shorter fields cannot.

Backend keywords: the invisible layer

The backend keyword field — Generic Keywords for non-media products, Subject Keywords for media — is the one surface no shopper and no assistant ever sees, and it is where one of the most expensive silent mistakes on Amazon lives. The field is capped at under 250 bytes, not characters. Most characters count as one byte, but accented and special characters count as two or more, and going a single byte over the limit can de-index the entire field. Spaces and punctuation are not counted toward the limit, and they are not needed — keywords are simply separated by spaces.

The rules are about coverage without waste: include synonyms, spelling variations, and abbreviations; use singular or plural but not both; skip stop words; and do not repeat words, include your own or competitor brand names, add ASINs, or use subjective terms like "best" or temporary ones like "new." Amazon reserves the right not to use every supplied term, and decides relevance with a machine-learning model that changes over time.

With the title now reserved for identity and Item Highlights carrying the visible searchable phrases, backend keywords return to their proper role: capturing the synonyms and variant spellings that have no place in customer-facing copy but still need to be indexed.

The fields as a system for AI readers

Read field by field, the listing looks like a set of separate boxes. Read the way an AI assistant reads it, it is a single structured record — and the assistant's decision to cite a product or skip it depends on how that record holds together.

The Genrise cluster describes capturing an AI citation properly as tracing it down through levels: brand, then product, then the specific claim the assistant made, then the evidence behind that claim, then the individual attribute on the page that supplied it. The full version of that capture-depth method, and the measurement hierarchy of share of conversation, share of answer, and share of voice, lives in the answer engine optimization piece. What the July change does is give brands a cleaner place to put the attributes and claims that capture depth traces — a searchable Item Highlights field sitting right under the title, exactly where an assistant looks for a product's distinguishing facts.

This is why the change matters beyond mobile readability. Adobe's 2026 analysis found that roughly a third of product-page content on major U.S. retail sites is effectively invisible to large language models — real, machine-readable gaps that keep products out of AI answers. Item Highlights is a new lever against exactly that invisibility, but only if it is built as structured claim-and-use-case content rather than filled with leftover adjectives. Used well, it feeds the same loop the rest of the cluster describes: the claims that win AI answers get identified, prioritized, and written into the field that indexes them — and the ones that don't earn their 125 characters get cut.

What changes on July 27 — and what to do first

The mechanics of the deadline are straightforward, and they reward leading rather than waiting. Until July 27, a brand can keep its existing title or move to the 75-character format and add Item Highlights at its own pace. After July 27, any title still over 75 characters will be updated to Amazon's AI-generated recommendation, gradually. Listings stay active throughout, and a brand can edit titles and Item Highlights at any time. Brand owners get a 14-day window in Review Listings Changes to review, modify, and approve the AI-generated recommendations before they go live.

The reason to lead is the Item Highlights opportunity. A brand that re-architects its own titles decides what stays in the identity line and what moves into the searchable highlights field. A brand that waits inherits Amazon's split — an AI rewrite that may put the right facts in the wrong field, or flatten the claim selection that took a category team months to get right. The 14-day review window is a safety net, not a strategy; reviewing thousands of AI-proposed rewrites under a clock is a worse position than authoring them deliberately in the weeks before the deadline.

The practical sequence for an enterprise catalog is to treat the window before July 27 as a focused sprint: bring titles to identity-only, build Item Highlights as prioritized claim-and-use-case strings, and let the fields below — bullets, description, backend keywords — settle into the roles above. The point is to arrive at the deadline having already made the decisions, rather than ratifying Amazon's.

Where this fits

The July change is a deadline, but the underlying shift is permanent: Amazon is restructuring the listing so that identity, searchable performance content, human conviction, context, and indexing each have their own field. That maps cleanly onto how Genrise grades content — a clean, identity-only title is Content Foundation; Item Highlights built as searchable claim-and-use-case strings is SEO Performance and AI Shelf Visibility working together; and deciding which claims earn the scarce 125 characters is Brand's Right to Win. The framework that grades a catalog against those dimensions is the PDP audit framework; this piece is the field-level architecture beneath it.

References

Amazon primary sources

• Amazon, Product title requirements and guidelines (Seller Central / Vendor Central Help). Character limit, information order, repetition, special-character, numeral, and capitalization rules; Item Highlights field, character count, comma-separated-phrase format, and search/PDP visibility; July 27, 2026 effective date and AI-recommendation behavior.
• Amazon, Product bullet point requirements (Help). Bullet count, length, formatting, and prohibited content.
• Amazon, Product description requirements (Help). Length, HTML, sentence case, prohibited content, A+ Content relationship.
• Amazon, Product keyword requirements (Help). Byte limit, generic vs. subject keywords, inclusion and exclusion rules.
• Amazon Vendor Central, Updates to improve your product titles begin on July 27 (seller/vendor announcement, June 2026).

Cluster-verified market data

• Amazon Q4 2025 earnings (reported February 2026): 300M+ Rufus customers in 2025; 149% YoY monthly-active-user growth; ~$12B incremental annualized sales.
• Adobe Analytics (2026): AI-driven retail traffic surge and conversion data; ~one-third of product-page content effectively invisible to large language models. business.adobe.com

Related Genrise insights

• What is digital shelf optimization
• Amazon Rufus deep-dive
• AI shopping assistants field guide
• Writing product descriptions for three audiences
• The enterprise PDP audit framework
• Answer engine optimization for commerce
• Product claims and AI visibility