A new SKU goes live Tuesday. Your photographer delivered images last week. The copy field is still empty. This tutorial replaces the “open Seller Central and stare at the title field” hour with a 90-minute workflow: 20 minutes of keyword and competitor research, 30 minutes drafting the title and five bullets, 20 minutes on the description, 20 minutes on A+ Content modules. It is aimed at private-label sellers, agency operators, and brand managers who launch 2-10 SKUs a month and cannot spend a four-hour copy session per product.
The difference between this and pasting “write an Amazon listing for X” into a chatbot is that here AI works inside Amazon’s actual 2026 field limits, against your real competitor set, and you keep final approval on every compliance-sensitive claim.
TL;DR
- Run the listing in five drafting steps: keyword clusters, title, five bullets, description, A+ modules, then backend search terms.
- Hard limits as of June 2026: title 200 characters (125 for apparel, 150 for electronics, 80 for baby and pet supplies); bullets 500 characters each for sellers but only the first 1,000 bytes across all five are indexed; description 2,000 characters plain text; backend search terms 249 bytes in US/UK/EU, 500 in Japan.
- Use Claude Opus 4.7 or Sonnet 4.6 for long-form bullets and description; use GPT-5.5 for keyword clustering and the byte-tight backend field.
- Premium A+ Content (seven modules) is free to Brand Registry sellers as of 2024, but you must publish a Brand Story and get five A+ projects approved first.
- After 30 days, pull Search Query Performance and feed real buyer terms back into the next refresh.
The 2026 field limits AI must respect
Every prompt below references these numbers. AI does not know them by default and will happily write a 240-character title that Amazon truncates. Paste the relevant limit into each prompt.
| Field | Limit (June 2026) | Notes |
|---|---|---|
| Title | 200 characters | 125 apparel/fashion, 150 electronics, 80 baby and pet supplies |
| Title mobile display | ~70-80 characters | Mobile is 70%+ of traffic; front-load the primary keyword |
| Bullet (each) | 500 characters (sellers); 255 (vendors) | Only the first 1,000 bytes across all five bullets are indexed for search |
| Product description | 2,000 characters | Plain text only; HTML removed July 2021 |
| Backend search terms | 249 bytes (US/UK/EU), 500 (JP), 200 (IN) | Byte-counted, not character-counted; one byte over silently de-indexes the whole field |
| A+ modules | 5 standard / 7 Premium A+ | Text overlays cap near 300 characters per module |
The byte trap matters most. Backend search terms count bytes under UTF-8: an ASCII letter is 1 byte, a German umlaut is 2, a Japanese character is 3, an emoji is 4+. Go one byte over 249 in the US and Amazon ignores the entire field with no error message. For EU and JP listings, count bytes, not characters.
Who this is for
Private-label sellers across US, EU, and JP marketplaces; Amazon agencies running 20+ client SKUs; brand managers launching seasonal lines; and resellers who need fast, search-optimized copy without hiring a freelancer per SKU. It works best when you already know your category and competitor set. It fails when you have never bought a competitor product yourself, because AI cannot manufacture product knowledge you do not have.
Before you start
- Buy your own product. Real ownership beats spec sheets; the copy is better when you describe what surprised you in the box.
- Pull the top five competitor ASINs in your subcategory. Screenshot their title, bullets, A+, and the 3-5 search terms they obviously rank for.
- Have your Helium 10, Jungle Scout, or DataDive export ready: keyword volume, relevance, and competition for at least 40 terms.
- Confirm Brand Registry status. Registered sellers get A+ Content and Premium A+; unregistered sellers get the description only.
- Document category compliance rules. Supplements, baby, and electronics each have specific claim restrictions, and AI will write a banned claim unless you tell it not to.
Step by step
- Keyword clustering. Prompt:
Here are 40 keywords with volume and relevance. Group into 4 clusters: primary intent, secondary intent, long-tail, and brand-adjacent. Then pick the single primary keyword that should anchor the title.Sanity-check the clusters before continuing. - Title draft. Prompt:
Write 5 title variants under 200 characters (under 125 if apparel). Front-load the primary keyword in the first 80 characters because mobile truncates the rest. Structure: brand, primary keyword, key feature, material or size, use case. No symbols, no all-caps words, no superlatives like "best" or "#1".Pick one and tune the cadence by hand. - Bullets. Prompt:
Write 5 bullets, each starting with a 2-4 word benefit in title case followed by a colon, then 1-2 sentences. Keep each under 250 characters so the most search-relevant words land in the first 1,000 indexed bytes. Cover: primary benefit, secondary benefit, durability or material, ease of use or setup, guarantee or support.Replace any vague claim with a number. - Description. Prompt:
Write a 200-word description in 2 short paragraphs, under 2,000 characters, plain text only. Open with the buyer's specific problem in their own words. Close with a concrete use-case scenario.This is the section AI floods with corporate filler; rewrite the closing scenario by hand. - A+ Content modules. Pick four modules: hero banner with headline, a comparison chart against your own catalog, a lifestyle image with a 3-bullet overlay, and a brand story. Prompt AI per module and paste in the ~300-character overlay cap so it does not overflow. If you have Premium A+ unlocked, add a video module and a fuller comparison module to fill all seven slots.
- Backend search terms. Prompt:
From the 4 keyword clusters, generate backend search terms under 249 bytes for a US listing — comma-separated, no words already in the title, no competitor brand names, no punctuation beyond commas.Count bytes before pasting into Seller Central; for a JP listing, change the cap to 500 bytes. - 30-day feedback loop. Pull Search Query Performance and Brand Analytics. Feed the actual buyer search terms back into step 1 for the next refresh.
How to unlock Premium A+ (worth the slot)
Premium A+ used to cost six figures a year; since 2024 it is free for Brand Registry sellers, and it gives you seven modules instead of five plus interactive formats like video, hotspots, and carousels. Eligibility is granted weekly (on Fridays) once you meet two conditions: publish a Brand Story on your Brand Registry ASINs, and get at least five A+ Content projects approved in the trailing 12 months. Qualify in one marketplace and you can publish Premium A+ in every country you sell in. If you launch more than a handful of SKUs, clearing this gate early pays back on every later listing.
Quality check
- Title carries the primary keyword in the first 80 characters; mobile truncates the rest.
- Each bullet leads with a benefit, not a feature. “Holds 32 oz” is a feature; “All-day hydration for 8-hour workdays” is a benefit followed by a number.
- No banned phrases. “FDA approved”, “best-selling”, “100% safe”, and category-specific bans must be absent.
- Description reads aloud without sounding like AI. Read it once; if it sounds like every other listing in the category, rewrite the opening sentence.
- A+ modules each do a distinct job. Two lifestyle modules carrying the same message waste real estate.
- Backend field is under 249 bytes (count bytes, not characters) with no words duplicated from the title.
Which model for which step
| Step | Recommended model | Why |
|---|---|---|
| Keyword clustering, backend terms | GPT-5.5 | Fast, follows hard byte/character caps reliably |
| Title variants | Either | Short task; pick the cleaner cadence |
| Bullets, description | Claude Opus 4.7 or Sonnet 4.6 | Less corporate filler in long-form benefit copy |
| Per-marketplace localization | See translation workflow below | Compliance claims need a review pass, not raw machine translation |
Most agencies run both vendors in the same workflow rather than committing to one. For a deeper comparison of the underlying assistants, see our ChatGPT vs Claude vs Gemini guide.
Common mistakes
- Treating AI bullets as final. Every AI-default bullet uses the same cadence; readers feel it and click away.
- Stuffing the title with every keyword. The algorithm rewards relevance, not volume; over-stuffed titles get suppressed.
- Writing claims AI cannot verify. “Clinically tested” without evidence gets the listing pulled.
- Skipping the buyer-search-term loop. The keywords that converted are rarely the ones you targeted at launch.
- Overflowing the backend byte limit. One byte over 249 in the US de-indexes the whole field silently.
- Reusing one A+ template across the catalog. Brand Registry rewards distinct modules; copy-paste modules waste the slot.
FAQ
- Which AI tool is best for Amazon copy?: Claude Opus 4.7 or Sonnet 4.6 for long-form bullets and the description; GPT-5.5 for keyword clustering and the byte-tight backend field. Most agencies use both in the same workflow.
- What is the title character limit in 2026?: 200 characters for most categories, but 125 for apparel and fashion, 150 for electronics, and 80 for baby and pet supplies. Mobile shows only the first 70-80 characters, so front-load the primary keyword.
- Why did my backend keywords stop working?: Almost always the byte limit. Backend search terms are capped at 249 bytes in the US/UK/EU (500 in Japan), counted in bytes not characters. One byte over and Amazon ignores the entire field with no warning.
- How often should I refresh a listing?: Every 90 days for active SKUs, every 30 days for SKUs in their first quarter. Refresh sooner if conversion drops 20% week over week.
- Is Premium A+ Content still free?: Yes as of June 2026. It is free to Brand Registry sellers who publish a Brand Story and get five A+ projects approved in the trailing 12 months. Amazon says it will notify sellers before reintroducing any fee.
- What if I sell in 5 marketplaces?: Draft in one language, then localize per market with the AI translation self-critique workflow. Never let AI machine-translate compliance claims without a review pass, and re-check byte counts in every market.
Related
- Ecommerce product description prompts
- AI content marketing stack guide
- AI translation self-critique workflow
- AI Content Series Planning Tutorial: 12-Episode Roadmap
- AI LinkedIn Thought-Leadership Post Tutorial Without Cringe
- AI TikTok Script Tutorial: Hook + Reveal + CTA in 60 Seconds
Tags: #Amazon #E-commerce #listing #Tutorial