Amazon shows search results as a grid or a list. Here is what that does to rank, attention and averages, and how to normalize positions across layouts.
"We rank 8th" sounds like a precise fact. On Amazon it is not, because the same position looks very different depending on how the results page is laid out. Position 8 on a multi-column grid sits near the top of the screen. Position 8 on a single-column list can be several scrolls down.
This guide explains the layouts Amazon uses, what attention research says about grids versus lists, why averaging raw positions across keywords misleads, and how to normalize positions so they stay comparable. Each section says whether a claim is documented by Amazon, described in research, alleged by a regulator, or only observed.
The two desktop layouts
Amazon does not publish a description of its search page layouts. What we know comes from observation, including systematic observation by journalists.
The grid. Most desktop searches return a multi-column grid of product tiles: image, title, rating, price and delivery promise, several tiles per row.
The list. Some searches return a single column with one product per row and wider tiles that show more text. The Markup, which collected desktop search results in January 2021 for its investigation of Amazon's own brands, reported that results were sometimes "delivered in a single column, one item per row," and that "this happened for some electronics searches but never in other search categories."
Sellers commonly observe the same split today: visually driven products tend to get a grid, while some spec-driven categories, electronics in particular, can get a list. Treat this as an observation, not a rule. Amazon can change the mapping at any time, and the layout for a given keyword can change too.
The split matches e-commerce UX advice: Baymard Institute recommends grid view "for visually driven products" and list view "for spec-heavy products." That explains the logic, but it does not prove how Amazon decides.
Screen width changes the grid itself. The Markup also noted that "Amazon changes the number of items across the first row based on screen size." A position that sits in the first row on a wide monitor can drop to the second row on a laptop. For that reason The Markup used the top-left spot as its main measure, writing that it "is the only one to remain the same across all search results in our data."
Layouts also show different numbers of results on page one. A list page and a grid page for two similar keywords can end at very different positions.
Mobile and the app are different again
The Markup states plainly that Amazon's search results "may differ on mobile, desktop, and the Amazon app," and it limited its own analysis to desktop.
The FTC's 2023 antitrust complaint against Amazon shows the difference with side-by-side examples. It describes a desktop results page where "no organic search results appear in the first row," and a mobile page where "the top two results are 'Sponsored' advertisements, and the third is a recommendation widget." It also alleges that for mobile shoppers, sponsored ads "are often the only results visible" without scrolling. These are allegations and illustrative examples in a pending case, not a measured average.
Sellers commonly observe that phones and the app show a narrower layout with fewer products per row than a desktop grid. Either way, a desktop position does not translate directly into what a phone shopper sees.
Sponsored results and widgets sit inside the page
Organic results share the page with ads and modules.
- Sponsored Products appear within the results. The FTC complaint describes them as ads "which appear within search results," while Sponsored Brands ads "appear above search results."
- Amazon Ads names the placements. Advertisers can adjust bids separately for "top of search (first page)," "rest of search" and product pages. So Amazon itself documents ad slots beyond the top of search.
- Research shows how they mix. A 2024 audit of Amazon desktop search (Dash et al.) found sponsored results shown "in the top, middle and at the end" of the results page. In its example, "the organic results start at the right hand corner of the top row."
- Widgets and carousels. Amazon also inserts modules such as editorial recommendations and themed carousels. The Markup excluded these from its main analysis because they "are not part of the grid." In a supplementary check, it found that counting sponsored and promotional carousels would lower the organic share in its measure from 87 to 68 percent.
The practical point: organic position 1 is rarely the first thing a shopper sees, and the gap between organic rank and screen position differs by keyword, layout and device.
What attention research says about grids and lists
There is no public eye-tracking study of Amazon's own result pages. The relevant evidence comes from web search, image search and other e-commerce sites.
Lists: attention falls steeply with position. Eye-tracking work on web search (Joachims et al., SIGIR 2005) showed that clicks are strongly biased toward higher-ranked results. Users click top results more often even when lower ones are more relevant, which the authors call trust bias.
The fold still matters. Nielsen Norman Group's 2018 eye-tracking study found that "users spent about 57% of their page-viewing time above the fold," and 74% in the first two screenfuls. This was general web browsing, not Amazon search, but it shows why "how far down the screen" matters as much as "what number."
Grids are not long lists. A WWW 2019 study of grid-based image search (Xie et al.) found three behaviors that list models miss: a middle bias (central positions in a row get more attention than the edges), slower decay (attention does not drop as sharply with rank), and row skipping (users sometimes jump past whole rows). Researchers working with Etsy (Guo et al., KDD 2020) built row-skipping and slower-decay click models specifically for grid-based product search, because list-based models did not fit. A 2025 study of a travel marketplace grid reported row-and-column effects "that simpler heuristics overlook."
Context changes the curve. Amazon researchers (Di Benedetto et al., 2023) note that examination probabilities likely depend on factors such as "the screen size," so the position-bias curve "depends on the context."
What about the F-pattern? Nielsen Norman Group identified the F-shaped scanning pattern in 2006. It describes how people scan text in a page's "content area," and NN/g calls it the default "when there are no strong cues." Product tiles with images, prices and badges are strong cues. Do not assume the F-pattern applies to a product grid.
What this means. A single "position value" curve borrowed from web search overstates how fast attention drops in a grid, and understates it in a single-column list. On Amazon, a shopper on a grid page sees many products at once and scans across them, while a shopper on a list page sees a few and scrolls.
Why averaging raw positions across layouts misleads
Suppose you track two keywords: "lunch box" returns a grid and "dash cam" returns a list. Your product sits at organic position 10 on both, and a tool reports an average rank of 10.
Those two 10s are not the same thing:
- On the grid, position 10 is in the upper part of the page and possibly above the fold on a wide screen.
- On the list, position 10 is several screens down and may be close to the end of page one.
Now suppose Amazon switches "dash cam" to a grid. Your position might move from 10 to 14 because more products fit on page one and ads are placed differently. The average says you got worse. Your visibility may have improved.
Raw averages fail in three ways:
- Unequal scales. Page one ends at different positions on different layouts, so the same number means a different share of the page.
- Hidden layout changes. A switch from list to grid or back moves every product's number at once, which looks like a ranking change.
- Off-page values. Products that drop off page one are often given an arbitrary placeholder rank, which skews any average.
The article on how many keywords to track explains why you should weight keywords by importance. Layout normalization comes before weighting. Otherwise you are weighting numbers that do not mean the same thing.
How to normalize positions
You cannot measure each shopper's real attention, but you can put every keyword on the same scale.
1. Relative position within page one. For each keyword and each capture, count the organic results on page one (call it N), then compute:
relative position = (position − 1) / (N − 1)
The top result scores 0, the last organic result on page one scores 1. Position 10 on a short list page and on a long grid page now get different values, which is correct.
2. A visibility curve that reaches zero at the end of page one. Convert relative position into a visibility score between 1 and 0, for example visibility = (1 − relative position)^k. With k = 1 the score falls in a straight line. With k above 1 it falls faster near the top, closer to the steep drop seen in list research. Anything off page one scores 0, so no placeholder ranks are needed.
This follows established practice in search evaluation. Discounted cumulative gain (Järvelin and Kekäläinen, 2002) discounts lower positions on a logarithmic curve, and rank-biased precision (Moffat and Zobel, 2008) models a user who continues to the next result with a fixed probability. Xie et al. adapted both to grids. Your choice of curve is a modeling assumption. State it and keep it fixed.
3. Use organic positions only. Count only organic results when computing N and positions. Ads change with bids. If you want to study ad pressure, track it as its own metric (for example, sponsored slots above your listing) instead of mixing it into rank.
4. Compare like with like. Desktop, mobile and app layouts differ. Capture from one setup, at one screen size, from one delivery address, so N and the layout are consistent over time.
5. Average visibility, not rank. Sum or weight-average the visibility scores across keywords. A product that sits halfway down page one on every keyword now gets the same value whether those pages are grids or lists.
How to read a rank change when the layout switches
Amazon tests page designs. In 2020, an observer of Amazon UI experiments documented a test that showed the same product tiles as a grid instead of a list on some search queries. That is one third-party observation, but it shows that a keyword's layout can change without notice.
When a keyword's numbers jump, check for a layout switch before blaming your listing:
- Did N change? A sudden change in the number of organic results on page one, on one date, is the clearest sign of a layout change.
- Did everyone move? If most products shifted by a similar proportion, the page changed, not the ranking. See the diagnostic steps in why your Amazon ranking dropped.
- Did relative position hold? If your raw position moved but your relative position barely changed, your standing is stable.
- Did the ad mix change? More sponsored slots or a new widget above you can lower your screen position without changing organic rank.
- Wait for several captures. Tests can flip back. Mark the date and judge the trend after the layout has settled.
Keep the history from before the switch. Relative scoring lets list-era and grid-era ranks sit on one line.
How JustRanks handles layouts
JustRanks records the whole organic first page of Amazon.com search results for more than 200,000 keywords over time, from a fixed US delivery address, overnight US time. Keywords saved in any user's market are captured every day, all others every 48h. Ads never count toward positions. Positions are scored on the whole first page of each keyword, so grid and list pages stay comparable, and Markets show visibility across up to 100 keywords per product. For background on how ranking itself works, see how Amazon organic ranking works and what a rank tracker measures.
Sources and further reading
- The Markup: How We Analyzed Amazon's Treatment of Its "Brands" in Search Results (2021)
- FTC: FTC Sues Amazon for Illegally Maintaining Monopoly Power (Sept. 26, 2023)
- Amazon Ads: Rest of search bid adjustment for Sponsored Products (2024)
- Dash et al., "Sponsored is the New Organic" (2024)
- Rock, Strauss, O'Reilly & Mazzucato, "Behind the clicks: Can Amazon allocate user attention as it pleases?" (Information Economics and Policy, 2024)
- Xie et al., "Grid-based Evaluation Metrics for Web Image Search" (WWW 2019)
- Guo et al., "Debiasing Grid-based Product Search in E-commerce" (KDD 2020)
- Di Benedetto et al., "Contextual position bias estimation using a single stochastic logging policy" (Amazon Science, 2023)
- Nielsen Norman Group: Scrolling and Attention
- Nielsen Norman Group: F-Shaped Pattern of Reading on the Web
- Joachims et al., "Accurately Interpreting Clickthrough Data as Implicit Feedback" (SIGIR 2005)
- Järvelin & Kekäläinen, "Cumulated Gain-Based Evaluation of IR Techniques" (ACM TOIS, 2002)
- Petrov et al., row and column effects in a travel marketplace grid (2025)