What the sources behind Perplexity's answers show, and what it means for your brand
A brand can vanish from AI recommendations for a week and come back the next. Between our March 20-21 and March 26 scans, six days apart, hiking boot brands moved noticeably in Perplexity's answers. None of them had changed their content. The sources Perplexity pointed to had changed. We believe Perplexity changed which sites it draws on.
Disclosure: Envoyra sells the AI visibility scanning tool used to collect this data. We have a financial interest in findings that show the value of ongoing monitoring. The method is set out in full below. Underlying scan data, including how often each source site was cited, is available to clients and press on request.
Background: four weeks of weekly scans
Since early March 2026 we have scanned a set of hiking boot brands on Perplexity Sonar every week. Each scan asks the same buyer questions about the same brands, the same way. Four scans were done by the time of writing:
Scan 1 (early March): first look. Partial coverage; several brands entered the data for the first time. Scan 2 (March 13-14): first scan covering every brand. Scan 3 (March 20-21): second full scan. Scan 4 (March 26): third full scan.
This article compares Scan 3 and Scan 4, the two latest full scans. Only five or six days separate them, the shortest gap in our data. That makes it the cleanest test of one question. Did results move because brands changed something, or because Perplexity did?
The finding in one sentence
A brand can vanish from AI recommendations for a week and come back the next, with no change to its website, content or marketing.
Between March 20-21 and March 26, about as many hiking boot brands gained ground on Perplexity Sonar as lost it. None of them changed anything we could see. What did change was which sources Perplexity cited, and we have that data.
How we ran the study
Platform: Perplexity Sonar (standard tier). Category: hiking boots. Scan 3: March 20-21, 2026. Scan 4: March 26, 2026. Gap: five to six days.
Each answer from Perplexity lists its sources. We call those citations. We saved them for both scans and compared which websites were cited, and how often.
The buyer questions were identical in both scans: same questions, same wording, same order. No question was added, removed or edited between scans.
What changed for brands
Between the two scans, some brands rose, a similar number fell and the rest held roughly steady. We saw no change on the brands' side.
Salomon, Oboz, LOWA, KEEN and Hoka gained ground. Oboz gained the most. Merrell and Columbia held steady. La Sportiva, Zamberlan, Scarpa and Danner lost ground.
Why we think this was Perplexity, not the brands
Four lines of evidence point to a change at Perplexity rather than changes by brands.
1. Movement was spread out, not isolated. If one brand changed its content, mainly that brand would move. Instead, many brands rose and many fell in the same week. A reshuffle across the whole category points to a change in where answers draw from.
2. The number of brands named per answer stayed flat. Perplexity was not naming more brands. It was naming different ones. That is swapping, not growth.
3. Salomon moved back and forth. Across three weekly scans, Salomon fell and then rose back. That looks like a brand moving around a stable level, not a brand that rewrote its content twice in three weeks.
4. Winners and losers split by the kind of sites that cover them. The brands that gained (Oboz, KEEN, LOWA, Salomon) are all well covered on US gear comparison sites. The brands that lost (Zamberlan, La Sportiva, Scarpa) are European specialty brands with less coverage on English-language review sites. Individual content changes would not produce that pattern. A shift in which sites Perplexity draws on would.
What the sources show
We compared how often each website was cited in the two scans. The total number of citations was almost the same. So Perplexity was swapping sources, not adding more.
Sites cited more often in the second scan: cleverhiker.com, myoutdoorbasecamp.com, backpackinglight.com, outdoorgearlab.com, and meindlusa.com, which was new.
Sites cited less often: switchbacktravel.com, mountaineerjourney.com, hikingfeet.com, bakershoe.com (which dropped out), rei.com and treelinereview.com.
The pattern: a move away from blog-style review sites (Switchback Travel, Mountaineer Journey, Hiking Feet) toward structured gear comparison sites (OutdoorGearLab, CleverHiker, Backpacking Light).
This was not what we first expected. Our first read was that REI had gained. The data shows REI was cited less often. The real shift was narrower: between these two scans, Perplexity cited opinion-led review blogs less and structured gear comparison sites more.
Why Oboz gained the most
Oboz's jump is the strongest sign that this was a change at Perplexity, not a content change.
Oboz did not update its website between the two scans. But Oboz is covered in depth on OutdoorGearLab and CleverHiker, two of the sites cited much more often in the second scan. When Perplexity leaned on those sites more, Oboz had the most to gain.
This is the clearest case in our data of a brand's AI visibility moving for reasons outside its control. A brand can have strong pages and still lose ground if the review sites that cover it fall out of favour. A brand with deep coverage on other sites can gain ground without doing anything.
If this is happening in your category, you'd never know without measuring. Explore the research →
How much results move week to week
Across four weekly scans, from early March to March 26, results in this category moved a lot on Perplexity Sonar. Most weekly moves were small. A few were very large.
We set aside moves where a brand was entering the data for the first time. Hoka is one example: it had no data in the first scan, so its first jump reflects entering the data, not real movement. The pattern held: mostly small moves, with a few very large ones.
La Sportiva shows real volatility. It was named in our first scan, disappeared completely the next week, then came back strongly in the March 13-14 scan. That was a move from a genuine zero, not a data-entry effect. A brand can vanish from AI recommendations for a week and come back the next.
Columbia was the most stable brand, nearly flat across all four weeks. Columbia and La Sportiva show the full range of movement in this category.
What this means for how you measure
You can't measure AI visibility with one snapshot. A single result could be a brand's low point, high point or normal level. Without weekly tracking, you can't tell which.
Find the cause before you act. A brand that sees a drop this week should not rush to rebuild its website. First ask what kind of drop it is. Was it the AI assistant changing its sources (outside your control, no action needed)? A rival improving its content (outside, needs a strategic answer)? Or a gap in your own content (inside, fixable)?
You need the source data to find the cause. Brand results show that something changed. The sources behind the answers show why. Without the comparison of cited websites in this analysis, we would have had a plausible guess (Oboz improved its content) that was wrong.
Fixed settings matter. In these scans we did not set Perplexity's randomness setting ourselves; we used its default. We have since fixed that setting in every scan. Results from late March onward are not affected by changes to it.
Limits
This analysis covers one AI assistant (Perplexity Sonar), one category (hiking boots) and a gap of five to six days. We can't confirm whether the same shift affected other categories or other Perplexity models. We inferred the change from shifts in which sites were cited. Perplexity did not announce any update in this period. The chain we describe (a change in sources, then a change in which brands are named) is our best explanation of the data, not a confirmed mechanism.
What this means for you
Measure quarterly and you can't see these swings. Measure monthly and you see them late. Measure weekly and you can tell three kinds of movement apart:
AI assistant change: the assistant shifted which sources it uses. Outside your control. No action needed. Rival gains: a rival improved its content or its coverage on other sites. Outside your control. Needs a strategic answer. Your own gap: something is missing from your content. Inside your control. Act on it.
Without weekly tracking and source data, all three look the same. A brand that sees a drop and rebuilds its website at once may be solving the wrong problem.
We track this weekly: which brands are named, shifts in the sites cited, which rivals are named in your place, and the most useful gaps to look at next.
Explore the research
Most brands assume they show up in AI recommendations. Many don't.
The research shows which brands are named, which are left out and how the two groups differ.
Data: March 20-26, 2026. Hiking boot buyer questions. Perplexity Sonar. Source analysis from stored answer data for both scans. Randomness setting not fixed in these scans; fixed from late March onward. The spread-out pattern of movement suggests that setting did not cause these changes. Envoyra Engine v1.0. envoyra.com. March 2026. Built by enterprise technology advisors with backgrounds in global digital strategy and technology sales.