16 brands · 3,224 report rows · 3,677 prompt clicks · 362 orders · June 6 to October 27, 2026 · Amazon US
01Amazon writes the prompts and bills them as ad clicks
Amazon announced Sponsored Prompts at its unBoxed conference in November 2025, as a free beta for US advertisers. Since March 25, 2026 they cost money. A prompt is a short line that Amazon's AI writes for one of your ads, usually a question about the product. It shows in search results and on product pages. Tapping it opens a conversation with Rufus, which Amazon has since made part of Alexa for Shopping, or answers the question on the page.
You don't set prompts up. Amazon adds them to every eligible Sponsored Products and Sponsored Brands campaign and uses the campaign's targeting. A prompt click is billed like any other click on that campaign. You can pause single prompts in the ads console, and you can't write your own. Amazon says it builds them from your product page, your Brand Store and your campaign data (Amazon Ads, March 2026).
Amazon offers a prompt report for the US only and keeps 95 days of it. Our app has saved it every day since August 13, 2026. Because of the 95 days, our data for most brands starts on July 12. For four brands it starts in May or June. The 16 brands sell things like insect repellent, personal care tools, batteries, cleaning supplies, towels, skin balm and hair care.
02Prompts took about 0.2% of the Sponsored Products budget
In the 13 weeks from July 27 to October 26, the median brand spent 0.22% of its Sponsored Products budget on prompt clicks. Across the 15 brands with prompt spend it ranged from 0.04% to 1.05%. Pooled, prompts were 0.35% of all their US ad spend.
That pooled number leaves out our biggest US account. It had no prompt click in 106 days, and neither did three smaller brands. The report only lists prompts that got a click, so we can't tell whether Amazon showed these brands no prompts, shoppers never tapped one, or the prompts were switched off. Our biggest account is 58% of the US spend, so with it the pooled share drops to 0.15%.
Prompts reached about a third of the products we advertised: 125 of 366 products with ad clicks in those weeks got at least one prompt click (34%).
The two brands with the most prompt spend sell personal care tools and insect repellent.
The share fell at the two brands with the most prompt spend: from 1.26% of Sponsored Products spend in August to 0.76% in October at one, and from 0.83% to 0.45% at the other. The other brands stayed between 0.1% and 0.3% in 13 of 15 weeks. One of the two sells insect repellent, a summer product. We can't tell from this data whether Amazon is showing fewer prompts.
03Prompt spend is already in your campaign reports
If you add prompt spend to campaign spend, you count prompts twice. The prompt report is a breakdown of clicks that are already in your normal reports. We checked all 2,306 campaign-days with a prompt click in Sponsored Products. The campaign report was never smaller than the prompt report. On the 73 days when every click on a campaign was a prompt click, the two matched to the cent. The advertised product report adds up to the campaign totals on 2,300 of those 2,306 days, so prompts are in there too.
Prompt clicks land in the rest of your data as well. They count toward the keyword or product target that served the ad, and the search term report lists them under what the shopper searched for. In the placement report, 57 of 69 prompt clicks on those prompt-only days were in Amazon's "Other on-Amazon" group, including all 30 three-cent clicks. 8 were at the top of search and 4 on product pages. At about 0.2% of spend, prompts move most keyword numbers very little.
04A prompt click brought fewer orders than an ad click, except one kind
To compare prompts with regular ads fairly, we matched them by product and week. For each product and week with prompt clicks, we took the same product's Sponsored Products clicks that week, with the prompt clicks taken out. Then we asked how many orders the prompt clicks would have brought at the ads' rate. Price, deals, stock and the season are the same on both sides of each pair. The shopper isn't: someone who taps a suggested question is at a different point than someone who clicks an ad.
Across 640 product-weeks (141 products, 15 brands), prompt clicks brought 0.53 times the orders of the same products' ad clicks. They cost 21% less per click. That average mixes very different prompts, so the chart splits it up.
Sponsored Products prompts with report dates up to October 18, so every order had its full 7 days to arrive. Sponsored Brands prompts name no product in the report, so they can't be matched this way.
Show the numbers
| Prompts | Product-weeks | Clicks | Orders | Orders per click | 95% range | Cost per click | Sales per $ |
|---|---|---|---|---|---|---|---|
| “Why choose [brand]?” | 205 | 466 | 138 | 1.63× | 1.25 to 1.86 | 0.02× | 74.12× |
| Other questions | 389 | 1,246 | 123 | 0.54× | 0.39 to 0.65 | 1.03× | 0.48× |
| Statements | 395 | 1,513 | 47 | 0.17× | 0.12 to 0.29 | 0.85× | 0.20× |
| All prompts together | 640 | 3,225 | 308 | 0.53× | 0.42 to 0.61 | 0.79× | 0.62× |
We grouped every prompt by its wording. “Why choose [brand]?” prompts brought more orders per click than the ads. Other prompts written as questions brought about half. Prompts written as statements, like a product name followed by what it's for, brought about a fifth. The next two sections take them in turn.
05“Why choose [brand]?” cost three cents a click
Some prompts ask why to choose the brand: "Why choose [brand] [product type]?". In Sponsored Products, Amazon billed 481 of 486 clicks on these prompts at exactly $0.03. That held across 10 brands, 104 campaigns and every month from June to October. On the same campaigns and days, regular clicks cost $2.20 on average (19,492 clicks).
The three cents shows up the same way in the campaign, product, keyword, placement and search term reports, which Amazon delivers separately, so it is the price Amazon charged. In Sponsored Brands the same kind of prompt was billed at a normal price, $0.45 to $1.85 a click depending on the month (279 clicks).
One click with no charge is left out.
These clicks also sold. Matched to the same product and week, they brought 1.63 times the orders per click of the ads (466 clicks, 138 orders, 10 brands, 95% range 1.25 to 1.86). They were 6% of all prompt spend and 48% of prompt sales.
Two things keep us from calling this free money. A shopper who taps “Why choose [brand]?” has already picked out the brand, so many of these orders might have come anyway. And one brand, which sells personal care tools, has more than half of these clicks. Without it the ratio is 1.42, and resampling whole brands gives a range of 0.95 to 1.74. So we only claim that they sold at least as well as the ads. Amazon doesn't explain the three cents. Our best guess is a minimum price when no other advertiser bids for that click.
06Questions brought half the orders of an ad click, statements a fifth
Every other prompt asks about the product or describes it. Some are questions, like "Does it [do the job] without [a common problem]?" or "Is it safe for [surface]?". Others are statements that read like a search phrase: "[Product] for [need]" or "[Product] with [feature]". The wording split the results more than the topic did.
Questions (1,246 clicks, 123 orders, 13 brands) brought 0.54 times the orders per click of the same products' ads (95% range 0.39 to 0.65). They cost the same per click, so they returned about half the sales per dollar (0.48 times). All 9 brands with at least 40 question clicks were below their ads (0.21 to 0.78 times).
Statements (1,513 clicks, 47 orders, 14 brands) brought 0.17 times the orders (95% range 0.12 to 0.29). They cost 15% less per click, which didn't make up for it: 0.20 times the sales per dollar. They took 49% of all prompt spend and brought 16% of prompt sales.
Show the numbers
| Brand | Question clicks | Orders | Vs ads | Statement clicks | Orders | Vs ads |
|---|---|---|---|---|---|---|
| Insect repellent | 368 | 39 | 0.50× | 394 | 8 | 0.10× |
| Personal care tools | 322 | 43 | 0.78× | 551 | 13 | 0.13× |
| Pet pads and towels | 169 | 13 | 0.55× | 220 | 8 | 0.28× |
| Skin balm | 70 | 6 | 0.44× | 107 | 13 | 0.65× |
| Household cleaning | 70 | 7 | 0.68× | 65 | 1 | 0.12× |
| Batteries | 62 | 6 | 0.27× | 54 | 1 | 0.07× |
| Home fragrance | 57 | 3 | 0.21× | under 40 | ||
| Hair care | 49 | 2 | 0.59× | under 40 | ||
| Body oil | 47 | 4 | 0.77× | under 40 |
Statements did worse than questions within each topic. For prompts about features, 10.3% of question clicks turned into an order against 1.9% of statement clicks. For safety and suitability it was 9.2% against 3.6%. They also did worse at 5 of 6 brands with enough clicks of both kinds. We can't see where Amazon shows each kind of prompt, so part of the gap may come from where they appear.
Without the brand with the most prompt clicks, questions come out at 0.46 and statements at 0.21. All prompts other than “Why choose [brand]?” together come out at 0.34 by product and week, 0.34 by product and day, and 0.36 when we match the same ad on the same day.
07Amazon's prompts ask about features and safety, almost never about price
We sorted all 546 prompts by topic. Most follow the product page closely: a repellent gets questions about its active ingredient, a towel about how well it absorbs. Only 2 invite a comparison, and none mention a gift, a price or a deal. A separate re-check sorted 150 of them without seeing our labels and agreed on 89%.
| Topic | Prompts | A typical one, rewritten |
|---|---|---|
| What it has or how it performs Does it [do the job] without [a common problem]? | 303 | Does it [do the job] without [a common problem]? |
| Is it safe or right for me Is it free of [ingredient]? Is it safe for [surface]? | 101 | Is it free of [ingredient]? Is it safe for [surface]? |
| A use or a problem it solves [Product] for [a specific problem] | 96 | [Product] for [a specific problem] |
| Why choose the brand Why choose [brand] [product type]? | 32 | Why choose [brand] [product type]? |
| What others say What do influencers say about [brand] [product]? | 12 | What do influencers say about [brand] [product]? |
| Comparison An invitation to compare the brand's products | 2 | An invitation to compare the brand's products |
| Gift, price or deal None | 0 | None |
All 546 unique prompts, one topic each. Examples are rewritten, never real prompt texts.
08What we can't tell from this data
- Click-through rate. The prompt report only lists a prompt on a day it got a click: 3,216 of 3,224 rows have one, and the rest have no impressions. Impressions from days without a click are missing, so we can't work out how often shoppers tap a prompt.
- New-to-brand. The Sponsored Products report has no new-to-brand columns. The Sponsored Brands rows show zero for all 42 orders, which looks like a missing field, so we report nothing.
- Credit for the sale. Amazon credits an order to the last ad click. If a shopper taps a prompt, chats, and then clicks a regular ad, the ad gets the order. Our ratios count orders credited to prompts, which may undercount what prompts added.
- Coverage. 16 brands, US only, mostly from July 12 to October 27. Amazon deletes the report after 95 days, so the first months after billing started are gone.
- Sponsored Brands. Their prompt rows name no product, so we compared them with the rest of the same campaign on the same day. With 271 clicks and 29 orders, 10.7% of prompt clicks turned into an order against 20.5% for the campaigns' other clicks. Too few to say more.
- Amazon's own numbers. Amazon says shoppers who tap a prompt convert 48% more often than shoppers who don't (September 2026), and that prompts on a Sponsored Brands ad raise its conversions by 6% (June 2026). Those compare shoppers who tap with shoppers who don't, or ads with and without prompts. We compare a prompt click with an ad click on the same product. They answer different questions.
09What we're doing in our accounts
- 01
Leave “Why choose [brand]?” prompts on. At three cents a click in Sponsored Products they cost almost nothing, and they sold at least as well as the ads.
- 02
Check the prompts that read like a search phrase. In the ads console's Prompts tab, look for prompts like "[product] for [need]". Ours took 49% of the prompt budget for 16% of prompt sales. Pause the ones with spend and no orders.
- 03
Judge question prompts like a weak keyword. They brought about half the orders of an ad click at the same price. Where your ads have margin to spare, that can still pay. Where they don't, pause them.
- 04
Don't add prompt spend to campaign spend. It is already in there, and in your keyword and search term numbers too.
- 05
Save the prompt report yourself. Amazon deletes it after 95 days. Download it every month, or pull it through the API every day.
- 06
Ignore the click-through rate in the prompt report. The report leaves out days without a click, so the rate it implies is too high.
10How we measured it
- Data. Amazon Ads' daily Sponsored Products and Sponsored Brands prompt reports, as our app stores them: 3,224 rows from June 6 to October 27, 2026, all US. Orders and sales use Amazon's 7-day attribution, the same as the Sponsored Products reports we compared them with.
- Coverage. Amazon keeps the report 95 days, and our app first pulled it on August 13, 2026, reaching back to different dates for different brands. We used each brand's data only from the first date we had pulled.
- Double counting. For every campaign and day with a Sponsored Products prompt click, we compared the prompt report with the campaign report and with the advertised product report.
- Matching. Product and week, Monday to Sunday, report dates up to October 18. The ad side is Amazon's advertised product report for the same product, all campaigns, with the prompt clicks taken out. A pair needs at least one click on each side. Expected orders are the prompt clicks times the product's ad conversion rate that week, and the ratio is actual over expected, summed over all pairs. Cost and sales work the same way.
- Ranges. The 95% ranges come from resampling products 20,000 times. Resampling whole brands, which is stricter, is in the tooltips and tables.
- Groups and topics. Every unique prompt text got one topic, by rules and then by hand. A prompt counts as a question when it ends in a question mark. “Why choose [brand]?” prompts are their own group.
- Spend share. Brand-days from July 27 to October 26 that the prompt report covers. Prompt cost divided by campaign cost.
- Re-check. A separate check recomputed every number on this page from the raw tables without seeing our code. It found two small errors, which are fixed. It also led us to split prompts by wording and to drop click-through rate.
- Limits. 16 brands we manage, US only, about four months. Two brands have most of the prompt clicks. The numbers describe what happened in these accounts. They don't prove what prompts cause.