Amazon Influencer videos and reviews

Amazon shut down Inspire, its TikTok-style discovery feed, in February 2025. Since then, creator videos live in three places: the “Videos for this product” block on a product detail page, search and browse placements, and the creator's own storefront. The detail page is the one that pays. A shopper who lands there has typed the product name or clicked a listing. They are not discovering; they are deciding.

So the video's job is narrow: answer “should I buy this one” with evidence the listing photos cannot give. Size against a hand. The noise it makes. How it opens. What the fabric does after a wash. Every craft rule below follows from that. Feed instincts (a scroll-stopping hook, a curiosity gap, a personality intro) are mostly wrong here, and the feed vocabulary (hook rate, saves, FYP) does not exist on this surface at all.

Storefront placement is automatic. Detail-page placement is earned. Amazon says this plainly: once a video is approved it appears on your Storefront, and from there it is consideredfor detail pages. Whether it gets there depends on three things — show the product genuinely in use, tell shoppers something they cannot get from the listing, and make it unmistakably clear which product the video is about. That last one is the sleeper. If you review a laptop but never say the brand or show the logo, Amazon cannot tell whether it is an Apple, an HP, or a Dell, so it may not reach any laptop's detail page at all.

A note on the three-video rule you may have read elsewhere: trade coverage reports that creators need three moderator-approved videos and onsite commission enrollment before videos reach detail pages, and that moderation takes about two business days. Amazon's own best-practices page does not mention any such threshold — it describes placement as earned per video, by content. Treat the count as unconfirmed, confirm it in your own dashboard, and optimize for the three criteria either way.

Two layers. Base commission by product category, reported in 2026 trade coverage at roughly 1% to 20% (fashion around 4% to 7%, beauty around 6% to 10%, some categories such as gift cards paying nothing). On top of that, an onsite bonus, reported at roughly 1% to 5%, on purchases made within 24 hours of a viewer watching your video on Amazon.

One change worth knowing: coverage of an April 2026 update says only the tagged product (or its variant) now counts toward the onsite calculation. Before that, other items in the same shopping session could credit to the creator. If your earnings model assumed cart-wide attribution, it is stale.

The larger income for active creators is reported to come from Creator Connections, Amazon's invite-only marketplace where brands post paid briefs (flat fees in the low hundreds to low thousands per post) layered on commission. That is closer to a brand deal than to affiliate mechanics, and it is where a portfolio of approved, well-made videos turns into leverage. None of these figures are guarantees; they are the bands the program has been reported to pay, and your category and volume decide where you land.

Amazon's quality review checks that a video is demonstrative (product in use) and descriptive. Creator guidance and the rejection reports agree on what passes:

• —Product on screen in the first three seconds. No intro card, no channel branding, no greeting. The shopper is already looking for the product; show it.
• —Hands on it.Open it, wear it, run it, pour it. A talking head describing a product that sits in the background is the “product not clearly shown” rejection.
• —Answer the page's question. Every listing has one or two things the photos cannot settle. Find them in the existing reviews and answer them on camera with something observable.
• —One honest concession. A review with no downside reads as an ad, and Amazon shoppers discount ads. “The lid is stiff for the first week” makes the rest of the video credible.
• —Film it like a customer. Natural light, handheld, real reaction. Ad polish signals sponsorship and lowers trust on this surface specifically.
• —30 seconds to about two minutes.Coverage cites a 30-second floor for placement eligibility; Amazon's own pages do not state one, so treat it as unconfirmed. Past two minutes, detail-page viewers leave. 1080p, MP4 or MOV. Landscape 16:9 fills the detail-page media block; vertical is accepted but plays smaller there.

The CTA is the action the viewer can already take on the page: add to cart, check the size chart, pick the two-pack. It is never “link in bio,” “swipe up,” or a website. That is not a style note; it is a rejection rule, and it is next.

Amazon tells creators a video was rejected and, often, only a category. These are the classes that recur across the program's content guidelines, its Community Guidelines, and 2026 rejection write-ups. Framing throughout is “likely to be rejected”: Amazon revises these without notice.

• —Prices, promotions, urgency.No prices, discounts, coupon codes, “on sale,” or time-sensitive claims. The video outlives the price.
• —External links.No URLs, no “go to xyz.com to learn more.” One exception, and it surprises people: promoting your own social channels is explicitly allowed. Facebook, Instagram, YouTube, Twitter, Pinterest. Amazon names them.
• —Unsupported actions.No asking for things the Amazon player cannot do: follow, subscribe, comment, turn on notifications, “check the description.” This is the one that kills repurposed YouTube and TikTok cuts. The words survive the re-upload even when the links don't, and an end card reading SUBSCRIBE trips it with nobody saying a word.
• —Customer reviews.Never ask viewers to leave a positive review.
• —Amazon marks. Their logos and service names need express written consent. An Amazon box used as set dressing counts.
• —Video text. Spelling and grammar mistakes — in the title, the description, or on screen — are a stated rejection cause. Proofread the lower-thirds.
• —Rights and infringement.No music you don't have rights to, no borrowed images or artwork. Amazon also warns against wearing clothes, jewelry, or accessories with prominent branding you don't have permission to show. Check what you're wearing before you hit record.
• —Personal information in frame.The most common unexplained rejection. A shipping or order label with your name or address, mail on the counter, a phone screen showing an order confirmation. Amazon's own rule names phone numbers and email addresses; the labels are how it usually happens. Amazon says it happened, not where. Check the unboxing frames first.
• —Health and medical claims. Claims about the body or a medical condition that are not accurate and substantiated. Saying the blender is fast is fine; saying the supplement fixed a condition is the rejection class.
• —Competitor and off-Amazon products. Comparing to a competitor sold on Amazon is allowed if it is factual and not derogatory. Referencing a product not sold on Amazon is likely to be rejected.
• —Missing disclosure. Gifted or brand-paid product is allowed with a clear, conspicuous disclosure on screen or spoken. Without it you have a program problem and an FTC endorsement problem at the same time.
• —Duplicate or copied content.The same video twice, a re-cut of the listing's own media, content near-identical to another creator's. Reposting a TikTok with the watermark still on is a frequent instance.
• —Product not clearly shown, poor quality, wrong tags. Product too small, out of focus, never demonstrated; audio unintelligible; a tagged product that does not appear in the video. Up to ten products can be tagged, but every one must be clearly featured.

The practical order: check the frames for personal information and the script for prices and links before you upload. Those three account for most rejections and all of them are fixable with a re-edit, not a reshoot.

The Influencer Program and the customer-review system are different surfaces with different rules, and creators cross them. A shoppable video can feature a gifted product with disclosure. A written customer review cannot be compensated at all: no free product, no discount, no refund, no gift card, no promised future consideration, from a seller or brand, in exchange for a review. Amazon banned incentivized reviews in 2016 and the only exception is its own Vine program, which labels every review it produces.

What makes a written review useful, and what Amazon surfaces: specific and product-focused (not the shipping experience), balanced rather than five stars on everything, with photos or video where they add evidence. Any connection between the reviewer and the seller has to be disclosed in the text. Reviews written as promotion, by the seller, an employee, a family member, or anyone with a financial interest, are removed and can take the account with them.

Creators see views and onsite-attributed purchases per video. That is the whole signal. There is no hook rate, no completion curve, no saves, no shares. Two directional reads are useful:

• —Views per approved video over timetells you whether the placement is live and whether the listing has traffic. It says nothing about the video's quality on its own.
• —Attributed purchases per view, compared across your own videos in the same category, is the closest thing to a creative signal. Read it as relative, never as an absolute conversion rate; the attribution window and the tagged-product-only rule both shape it.

Importing feed-video vocabulary into an Amazon report invents data. State the two numbers the program gives you, state that they are directional, and make creative decisions by comparing your videos against each other.

Upload the video file or paste the script (Amazon detail-page video links cannot be fetched, so use the file). The classifier detects the format from a script; API callers can set the medium override explicitly. The analyzer then scores against a dedicated Amazon Influencer rubric rather than feed criteria. The rubric's primary driver is purchase-decision confidence on the detail page, and it puts moderation risk ahead of craft, because a rejected video earns nothing.

Concretely, the review checks that the product is on screen and in use within the first three seconds, that the page's real question gets an observable answer, that there is at least one honest concession, and that the CTA is an on-page action. It flags every rejection class above as a policy note or platform-policy risk flag and makes the first one found the biggest lever: prices and promos, off-Amazon links, personal information visible in frame (with the timestamp), unsubstantiated health claims, off-Amazon competitor references, missing gifted-product disclosure, duplicate content, product not clearly shown, and tags that do not match. On measurement it speaks the program's own language: views and attributed purchases, stated as directional, with no invented click-through rate and no feed jargon.

What the score can tell you: whether the video will clear moderation and whether it does the detail-page job. What it cannot tell you: how many purchases it will attribute. See the scoring rubric for how the score is built, and commercial content disclosure for the disclosure rules that apply across platforms.

The Ad Bench methodology: The checks and rejection-risk flags in this article reflect The Ad Bench's current Amazon Influencer video scoring rubric and creative-review approach. They are designed to identify moderation and creative risk before you upload, not predict or guarantee views, placement, or commission.

1. Amazon, Influencer Content Guidelines: Video Shorts (amazon.com/b?node=21423477011) and Influencer Best Practices: Video Shorts (amazon.com/b?node=21423479011). Guideline text transcribed from the program's published Video Content Guidelines and its “Best Practices: Shoppable Videos for Detail Page” page, both 2026-09-22. The pages were not fetched from the authoring session; verify against the live pages, which Amazon revises without notice.
2. Amazon, Community Guidelines (amazon.com/gp/help/customer/display.html?nodeId=GLHXEX85MENUE4XF) and Anti-Manipulation Policy for Customer Reviews. Cited by title; verify against the live pages.
3. TechCrunch, “Amazon kills Inspire, its TikTok-style feed for discovering products,” February 18, 2025.
4. Miles Insights, “Amazon Influencer Video Rejected? Every Reason It Happens in 2026” and “Rejected for Personal Information Visible? Here's the Fix,” 2026. Rejection classes and personal-information examples.
5. Influencer Marketing Hub, “Amazon Creator Hub 2026 Guide” and “Average Amazon Influencer Rates Explained,” 2026. Commission bands, onsite bonus, Creator Connections. Directional, not verified against Amazon's rate card.
6. Hubfluence and Referazon, Amazon Influencer Program 2026 requirement and commission guides. April 2026 tagged-product-only attribution change. Directional.
7. The Ad Bench methodology, Amazon Influencer video rubric (lib/claude/format-rubrics.ts). Reviewed September 17, 2026.