1Overview
Prepared alongside the article AI Product Recommendations: Why Your Brand Isn’t in the Answer. This multi-tab document works the full LSX Partners AI visibility process through on a real brand: gathering audience intelligence, building a prompt audit, measuring where the brand appears in AI answers, and setting out a strategy to improve that inside the growth areas the company has named for itself. Executed as an educational exercise, this AI visibility audit and strategy was prepared using all publicly sourced information, with no internal conversations or data from the Columbia Sportswear brand.
This icon marks a strategic read — my interpretation of the measurement beside it, and what I would do about it.
AI visibility baseline
Every engine describes the brand consistently and correctly — accessible, functional, good value, strong in rain and cold. Asked directly whether Columbia is a budget brand, all six answered no. No misinformation appeared anywhere in the captured set.
On unbranded questions built around the categories Columbia has stated it intends to own, the brand is named 25.5% of the time. Three quarters of the time a buyer asks AI about one of those categories, Columbia is not in the answer at all.
Buyers rarely open with a brand. They describe a situation — a first hike in the fall, standing outside in the cold, a long day on the water — and ask what to buy.
L.E.K. Consulting found that 53% of AI shopping use is comparing options and narrowing a shortlist, ahead of both discovery (19%) and final decision-making (21%). That middle stage is where a brand either enters the consideration set or never does.
The categories tested are the five growth pillars Columbia named on its Q2 2026 earnings call (August 2026).
AI audit highlights
The figures a marketing lead is most likely to be asked about, from AI answers captured across six engines on September 1, 2026.
Why most of the strategy recommendations sit off the brand’s own site
Columbia's own site presents this material correctly: navigation is organized by the activities buyers search — Hiking, Trail Running, Sun Protection, PFG Fishing, Ski & Snow — and product pages carry weight in ounces and grams, lug height in millimeters, and named technologies with definitions.
AI answers give a brand's own site no special weight. They are assembled from whatever is publicly retrievable across the web, and the brand site counts as one source among many. A marketing team can be executing correctly against the growth pillars on-site and still be absent from the answers.
Most of what follows is off-site: third-party review coverage, retailer product data, and the community sources engines cite.
Two of the five recommendations are on-site work. The other three sit in third-party channels — retailer listings, review coverage, community threads — which the brand can influence but cannot publish to directly.
Method
| Pass | Output |
|---|---|
| Audience intelligence | Buyer language from outdoor communities, used to build the prompt set |
| Prompt design | 25 buying questions — 18 unbranded, 7 naming Columbia directly — tagged by pillar, buyer type and journey stage |
| Engine audit | 150 runs across six engines; 5 returned no AI answer, leaving 145 captured, timestamped, with cited sources and screenshots |
| Competitive read | Every brand named in every answer, tallied by category |
| Agentic readiness | Crawler policy, retrievability, live price and availability |
Six engines: ChatGPT, Perplexity, Google AI Overview, Gemini, Claude, Bing.
Run September 1, 2026 from a fixed US location, logged out, memory and custom instructions off, one fresh session per prompt.
Columbia is not a client. This is a demonstration set, run on public sources to outline the LSX Partners AI visibility process.
Sources
| Finding | Source |
|---|---|
| Growth pillars, Q2 2026 segment detail | Columbia Sportswear Q2 2026 earnings call and investor release, August 3, 2026 |
| AI visibility figures | 25 prompts × 6 engines, 145 captured answers, September 1, 2026, fixed US location, clean sessions |
| Buyer language and category themes | Outdoor communities on Reddit, captured August 2026 |
| Organic ranking comparison | Semrush domain_organic, columbia.com and merrell.com filtered to hiking boot keywords, US database, September 2026 |
| Technology definitions | columbia.com/c/technology-omnitech/; columbia.com/tech-dry.html; September 2026 |
| Crawler policy, retrievability, price | columbia.com style 1594732; columbia.com/robots.txt; competitor robots.txt files; September 2026 |
| Agentic commerce forecast | Bain & Company, 2030 Forecast: How Agentic AI Will Reshape US Retail, December 17, 2025 |
| AI shopping behavior | Alchemer 2026 Retail Report (1,002 US shoppers); L.E.K. Consulting (2,650 US consumers, April 2026) |
Tags used in this document
Broken, and blocking other work.
Net-new work.
A position worth protecting.
2Visibility baseline
Prompts spanning the five stated growth categories and the full buying journey, run across six AI engines.
2.1 Performance against stated growth pillars
| Stated growth pillar | Category tested | Named in |
|---|---|---|
| 1Own the trail | Hiking boots and shoes | 11.1% |
| 2Dominate warmth | Insulated outerwear | 37.5% |
| 3Power PFG | Performance Fishing Gear | 43.5% |
| 4Fuel outdoor lifestyle | Casual apparel, fleece | 40.0% |
| 5Accelerate footwear | Hiking boots and shoes | 11.1% |
Hover a pillar for what it covers. Pillars 1 and 5 both map to hiking footwear, so they share a figure. Stated by Tim Boyle, Chairman and CEO, Q2 2026 earnings call, August 3, 2026.
Pillars 1 and 5 return the lowest figures. Pillar 3 returns the highest, and sits in the territory where competitors are least often named.
Put plainly: the two categories Columbia named first and last are the two AI mentions it in least.
2.2 Performance by engine
| Engine | Named in | Underlying index |
|---|---|---|
| Google AI Overview | 33.3% | |
| Gemini | 33.3% | |
| Perplexity | 27.8% | Own crawler, blended |
| Bing (AI answer) | 25.0% | Bing |
| ChatGPT | 16.7% | Bing |
| Claude | 16.7% | Brave |
Bing returned no AI answer on 5 of its 25 prompts, including both transactional ones, which is why 145 answers were captured from 150 runs.
Traditional SEO does carry into AI visibility. Columbia’s Google rankings are most likely why the two Google-index engines name the brand twice as often as ChatGPT and Claude.
Against Merrell those rankings are close: the two split the unbranded head terms, Columbia taking both men’s and Merrell both women’s.
| Same unbranded keyword | Volume | Columbia | Merrell |
|---|---|---|---|
| hiking boots for men | 12,100 | 1 | 3 |
| men’s hiking boots | 8,100 | 1 | 3 |
| women’s hiking boots | 9,900 | 4 | 1 |
| womens hiking boots | 9,900 | 3 | 1 |
| hiking boots waterproof | 110,000 | 14 | 6 |
| Named in hiking footwear AI answers | — | 11% | 48% |
Google organic positions, Semrush US database, September 2026. Ranking level with a competitor while being named a quarter as often is the gap 5.1 and 5.5 set out to close.
2.3 How the engines characterize the brand
Prompts that named Columbia directly and asked what the brand is and what it is worth.
“best known for affordable, practical outdoor clothing”
“durable, affordable outdoor apparel”
No, in all six. Mid-tier, value-oriented.
“worth it, especially if you buy it on sale”
- All six engines describe the brand the same way, and accurately: mid-tier, value-oriented.
- That framing is where the brand already wins: “Best hiking boots under $150” returns Columbia in four of six engines, against 11.1% across hiking footwear overall.
- Four of Columbia’s six hiking footwear appearances come from that one price-anchored question. Strip price out of the prompt and the brand largely disappears.
Recommendation: decide whether to own the value position or contest the performance one, then brief content and PR against that single answer. The pillars say “own” and “dominate”; the engines say value. Both can be true, but only one can lead.
2.4 Competitive share of hiking footwear answers
| Brand | Named in | Mentions | Share |
|---|---|---|---|
| Merrell | 48.1% | 26 | 22.4% |
| Salomon | 33.3% | 18 | 15.5% |
| Keen | 22.2% | 12 | 10.3% |
| Altra | 20.4% | 11 | 9.5% |
| Lowa | 20.4% | 11 | 9.5% |
| Hoka | 13.0% | 7 | 6.0% |
| Columbia | 11.1% | 6 | 5.2% |
| Other brands | — | 25 | 21.6% |
Two measures. “Named in” is the share of the 54 answers that mention the brand; an answer names several brands, so it does not sum to 100%. The chart divides the 116 total mentions, which does.
- Merrell is named in nearly half these answers. Columbia is named in one of nine, behind five footwear specialists and Hoka.
- Columbia holds 5.2% of all brand mentions. Another 21.6% is spread across eight brands outside the top six. No single brand holds the category.
Section 4 sets out which product data an assistant can and cannot retrieve from Columbia, and how that compares with the three brands above it. 5.1 is the fix.
3Audience intelligence
What buyers say when they talk about this category, and where Columbia sits in that conversation.
149 threads from 10 outdoor communities, captured August 2026, running back about four years. The prompt set was built from this data.
The buyer conversation ranks the brand the same way the engines do.
Fishing is where Columbia comes up most in real buyer threads.
Hiking footwear is where it comes up least.
Both data sets put fishing first and hiking footwear last.
3.1 What buyers weigh, by categoryfootwearouterwearfeeds 5.5
65 of the 149 threads are mostly about footwear, 81 mostly about outerwear and apparel.
| Reason given | Footwear 65 threads | Outerwear 81 threads |
|---|---|---|
| Durability, lasted N years | 672 | 528 |
| Waterproof performance | 604 | 589 |
| Fit and sizing | 509 | 302 |
| Comfort, no break-in | 452 | 173 |
| Weight | 410 | 796 |
| Price, or on sale | 376 | 1038 |
| Warranty or repair | 118 | 175 |
Mentions of each term family within each group of threads. A thread can carry several.
- In footwear, durability and waterproof performance lead, with fit close behind. How long the boot lasted is the most common reason given.
- In outerwear, price leads by a wide margin — nearly double the next reason, and 2.8 times what price scores in footwear.
- Warranty and repair is last in both, and in these threads that ground belongs to Patagonia, REI and Darn Tough. Columbia’s warranty does not come up.
Recommendation: Owned — brief footwear content on durability, dryness and fit, and outerwear on price and warmth. Third-party — push the same attributes into retailer listings and review coverage, where most of these answers get assembled. Detail in 5.5.
3.2 What the category is arguing aboutfootwearfeeds 5.4
| Theme | Mentions |
|---|---|
| Trail runners, versus boots | 496 |
| Waterproof and Gore-Tex | 395 |
| Merino and wool socks | 289 |
| Ankle support | 128 |
| Break-in period | 64 |
| Wide toe box, wide feet | 54 |
Trail runners is the single largest theme in the data set.
710 mentions of buying used or discounted: REI 465 including garage sale and ReSupply, Amazon 101, Costco 72, outlet 48, Sierra Trading Post 22.
In warmth threads buyers group Columbia with Costco and Eddie Bauer. The value position shows up in which brands it gets listed beside, and not only in the words used to describe it.
The most common subject in the buyer data set was whether to buy boots at all, versus trail runners — lighter, more comfortable, often cheaper, no break-in period.
Columbia sells trail running shoes. The Konos product line was named on the Q2 2026 call as part of the Omni-Max technical footwear range, alongside Tellurax and Peakfreak in hiking.
Named in 0 of 12 answers to “trail runners or hiking boots” and “trail runners with bad ankles” — the two prompts where that decision is made.
- The pillar covers hike and trail run. The audit found the brand in neither conversation.
- The product range already covers both. The gap sits in how the trail-running line is described publicly.
Recommendation: describe the Konos line by name in trail-running content and structured data, with the attributes buyers compare — weight, drop, lug depth, cushioning. An engine can only recommend a product line it has seen described.
3.3 The attributes buyers decide on
The attributes buyers weigh in the data set are the same fields an AI agent filters on.
| Attribute buyers weigh | Agent can filter | Published by Columbia |
|---|---|---|
| Waterproof | Yes | Proprietary name only |
| Price | Yes | Inconsistent — see §4 |
| Named membrane | Yes | Proprietary name only |
| Ankle height | Yes | Implied by product name |
| Weight | Yes | Published — 15.9 oz / 452 g |
| Toe-box width | Yes | Published — Wide sizing |
| Break-in required | Hard | Not published |
| Resoleable / warranty | Yes | Not published |
- The product pages carry more attribute detail than most brands publish: weight to a tenth of an ounce, lug height in millimeters, technologies named and defined.
- Only 2 of the 8 attributes are published in a form an agent can read. The other six sit behind a proprietary name, a product name, or are missing.
Recommendation: publish the missing attributes as labeled, machine-readable fields, so each value sits in its own field instead of inside a sentence in the product description. Section 4 covers why agents are not reaching even the ones that exist.
3.4 Proprietary vocabulary, by community tractionall categoriesfeeds 5.3
| Term | Mentions |
|---|---|
| PFG | 52 |
| UPF generic equivalent | 45 |
| Silver Ridge product line | 35 |
| Omni-Heat | 28 |
| Omni-Tech | 8 |
| Omni-Shade | 5 |
Semrush shows Omni-Shade at 320 searches a month. The community says it five times. Search volume and community vocabulary are separate measures.
- When buyers talk about waterproofing they say Gore-Tex. It comes up 580 times. Columbia’s own Omni-Tech comes up 8.
- PFG and Omni-Heat have made it into how buyers talk. Omni-Tech and Omni-Shade have not.
- Silver Ridge, a product name rather than a technology, is said more often than both of them combined.
- An engine answering “waterproof hiking boot” has to connect that phrase to Omni-Tech on its own. Where the connection is not published, it does not happen.
Recommendation: Owned — publish the generic equivalent beside each trademark. Third-party — get the same pairing into retailer listings and review coverage, which is where the unbranded question gets answered. Both are set out in 5.3 and 5.5.
3.5 Which brands buyers raise on their ownfeeds 2.4
| Brand | Mentions | Threads of 120 |
|---|---|---|
| REI | 397 | 76 |
| Patagonia | 361 | 51 |
| Salomon | 182 | 42 |
| Merrell | 268 | 39 |
| Keen | 128 | 30 |
| Lowa | 106 | 30 |
| Oboz | 51 | 20 |
| Arc’teryx | 31 | 18 |
| Darn Tough | 49 | 17 |
The brands buyers bring up themselves, without being prompted about Columbia.
REI appears in more threads than any brand, acting as retailer, house brand and advice source at once. Columbia’s own standing is measured in 2.1 and 2.4.
3.6 Verbatimfeeds the prompt set
Verbatim from customer and category conversations.
A prompt set is a living list. It should grow and change as new buyer data comes in.
This one was built from the threads above, which is why it covers the arguments buyers are actually having: trail runners versus boots, fit and sizing, and buying secondhand. A set drawn from a brand’s own category list would have carried none of them.
Recommendation: treat audience intelligence as a standing effort. Buyer questions shift, so re-capture on the same cadence as the prompt run.
4Product data readiness
Being recommended is one half of AI visibility and optimization. For a brand selling through retailers, the other half is whether an agent can reach the product, read stock and size, and return the price the buyer will actually pay.
Four assistants quoted four different prices for the same boot on the same day.
Verified price on columbia.com, style 1594732, same day: $110.00
4.1 Live price and availability test fix
“Is the Columbia Newton Ridge Plus II available in a men's 11 wide right now?”
| Engine | Price quoted | Sourced from |
|---|---|---|
| Google AI Overview | ~$110 | Backcountry, DICK'S |
| ChatGPT | $109.99 | Backcountry |
| Gemini | $77.00 | lenonlures.com, BeyondStyle |
| Perplexity | $50.00 | attributed to columbia.com |
| Claude | not stated | Columbia |
| Bing | no AI answer | — |
An assistant quotes the price it finds on a page it can read. The two that read authorized retailer listings matched the verified price. Perplexity named $50, attributed it to columbia.com — which showed $110.00 that day — and stated in the same answer that it could not read that page’s live inventory.
The reachability problem is the one Columbia controls, and 4.2 is where it sits.
4.2 Why an assistant cannot confirm stock and size fix
Columbia’s product pages and sitemap index both return normally to a browser, to Googlebot and to no user-agent at all. The problem sits in two lines of the robots.txt file.
1 · The endpoints holding stock and size are disallowed
robots.txt is a public file at the root of every website. It lists which addresses crawlers may and may not fetch.
Disallow means do not fetch this. Below are the live contents of columbia.com/robots.txt, retrieved September 2026.
Disallow: /Product-GetAvailabilitystock by sizeDisallow: /Product-Variation*size and color optionsDisallow: /Product-AllSizeSearch*size availability lookupDisallow: /Product-Detailproduct detail controllerDisallow: /Product-Showproduct page controllerDisallow: /search?cgid=*category browseUser-agent: ClaudeBotthe one AI crawler namedCrawl-delay: 1slowed, not blockedStock by size and the size and color options are the exact fields needed to answer the test question in 4.1 — whether one boot is available in a men’s 11 wide. Both are closed.
2 · Retailers supply the part that is missing, and they disagree
Unable to read stock and size from Columbia, the assistants answered from whichever seller they could reach. Backcountry, DICK’S and Columbia’s own store appear as sources, and so do lenonlures.com and BeyondStyle — third-party sellers carrying the brand name. Across those listings the same boot was quoted at $77.00 and at about $110.
Every price in the finding above came from this step. When the authoritative answer is unreachable, the assistant substitutes the best source it can get to, and the quality of that substitute is outside Columbia’s control.
4.3 Competitor comparison fix
| Brand | Availability endpoint | AI crawlers named in robots.txt | Share of trail answers |
|---|---|---|---|
| Merrell | Open | 8 named, each allowed | 48% |
| Salomon | Open | none named | 33% |
| Keen | Open | none named | 22% |
| Columbia | Disallowed | 1 named, crawl-delay only | 11% |
Columbia is the only one of the four that stops AI assistants from checking whether a boot is in stock in the size the shopper wants. It is also the only one below 20%.
Merrell goes furthest and lists eight AI systems by name as welcome. Worth copying, though it does not explain Merrell’s position — Salomon and Keen name none and still rank second and third.
Recommendation: let the assistants check stock and size (5.1), and add the welcome list while you are in the file.
5Strategy & priorities
Five recommendations, in priority order.
5.1 Let the agents in fix
Let assistants read stock and size by opening /Product-GetAvailability and /Product-Variation*, and name the AI crawlers explicitly instead of leaving them to the catch-all rule.
Both are edits to robots.txt. The team that owns that file can make them today.
5.2 Govern product data across the network build
Read back what every retailer has actually published for each product, compare it against the record Columbia sent them, and correct the gaps on a schedule.
Columbia already runs a Product Information Management (PIM) system, pushing records out to its retailer partners. Governing product data now reaches past that system: what each retailer publishes after mapping the feed into its own fields, and what an assistant quotes back from those pages, sits outside what a PIM covers.
Most product data programs stop at the feed. The answer a shopper gets is assembled after that, from pages Columbia does not control.
How the read-back runs:
- Take the products carried by three or more retailers. Those are where an assistant finds the most conflicting versions.
- For each one, capture what every retailer’s live listing publishes: price, availability, size and width options, and the decision attributes from 3.3.
- Compare that against the record sent from the PIM and log each difference by retailer and by field.
- Ask the assistants the same product question and record what they quote, so the answer a shopper receives is measured alongside the listings that produced it.
- Work the log by size of gap. Re-run price and availability weekly, specs and attributes monthly.
The differences come from a handful of predictable places:
- Retailers run on their own update schedules, so the same product refreshes at different times.
- Each retailer re-maps the feed into its own fields. Attributes with no match get renamed or dropped.
- Retailers set their own promotional calendars, so one discounts a product while another holds full price.
- Marketplace sellers list old stock numbers and discontinued colorways that no longer come from the feed.
- Merchandising owns the attributes, channel owns the retailer feeds, e-commerce owns the site. Where those teams work separately the versions drift apart, which is what the price test picked up. It is a common pattern in brands that sell mostly through wholesale.
Start with the products carried by three or more retailers. Those are the ones an assistant is most likely to find several conflicting versions of.
Bain forecasts US agentic commerce at $300–500 billion by 2030, roughly 15–25% of all e-commerce. Today an assistant quotes a wrong price and a person catches it at checkout. As agents move from recommending to transacting, that check disappears.
Adoption is expected to run faster for spec-driven buys like batteries than for considered ones like apparel and travel, which puts footwear on the slower end of that curve. That works in Columbia’s favor: opening retrievability and governing product data across the retailer network takes months, and the category timeline leaves room to do it properly before agents are completing these purchases at volume.
5.3 Teach the proprietary names buildevidence 3.4
The on-site work is already done. Columbia publishes a dedicated technology page at columbia.com/c/technology-omnitech/ describing Omni-Tech as waterproof, breathable and fully seam sealed. The engines have read it: asked directly what Omni-Tech is, Google AI Overview volunteers that it means waterproof breathability.
The explanation does not travel. Omni-Tech is near-absent from buyer language in the data set, where people say “waterproof” and never the trademark. The connection also runs one way only: an engine asked what Omni-Tech is can answer, and an engine asked for a waterproof hiking boot with no brand named does not arrive at Columbia.
The work is off-site. It means getting the same explanation into the sources an engine assembles an unbranded answer from — retailer listings, review coverage, community threads — so a shopper who describes wet feet on a cold hike, naming no brand and no technology, is answered with the product built for it. That is the program in 5.5.
5.4 Defend PFG before contesting trail defendevidence 2.1
PFG, Columbia’s Performance Fishing Gear line, is the strongest position in the audit and the least contested: named in 43.5% of answers, with almost no rival brands named alongside it.
The work is to be the brand named whenever someone asks about sun protection on the water, in the words buyers actually use as well as the branded ones. UPF, the sun-protection rating printed on fabric, comes up 45 times in the buyer data against 52 for PFG itself, so both vocabularies need answering.
With no incumbent to displace, this is the least expensive item on the list, and it can be secured while the footwear work is still being scoped.
5.5 Answer the decision attributes where the engines already read buildevidence 3.1
Weight, ankle support, break-in period and toe-box width are what buyers weigh, and the data already exists on Columbia’s product pages. 5.1 through 5.4 get that data published and readable on the brand’s own property. 5.5 puts the same answers into the retailer listings, review coverage and community threads the engines quote.
The answer a shopper receives is assembled from third parties, and Columbia does not publish in any of them.
Retailer listings — the brand supplies this already
Backcountry, DICK’S and Columbia’s own store appear as cited sources in the captured answers, alongside third-party sellers Columbia does not feed. The authorized listings are the one third-party surface it already controls. Extend the feed from price and stock to the four decision attributes, and hold retailers to them the way a brand holds them to imagery standards.
Owned by: channel and e-commerce. Starts with: the products carried by three or more retailers.
Review and comparison publishers
Engines answering these prompts pull from tested roundups. Running three of the audit’s hiking-footwear prompts as searches in September 2026 returns the same publications repeatedly:
| Publication | Appeared in |
|---|---|
| Mountaineer Journey | 3 of 3 |
| Switchback Travel | 2 of 3 |
| Treeline Review | 2 of 3 |
| Better Trail | 2 of 3 |
| OutdoorGearLab | 1 of 3 |
| GearJunkie | 1 of 3 |
| RunRepeat | 1 of 3 |
| REI Expert Advice | 1 of 3 |
| CNN Underscored | 1 of 3 |
| TGO Magazine | 1 of 3 |
Three prompts searched: wide feet, no break-in, lightest waterproof. September 2026.
The boots these roundups name are Merrell Moab 3, Altra, KEEN Targhee, Oboz Bridger, Lowa Renegade and Topo — the same brands the engines return, in roughly the same order.
These publications work in different ways. OutdoorGearLab states that it buys every product it reviews at retail and refuses free evaluation units. Treeline Review describes its picks as independently selected by its editors. REI Expert Advice sits inside a retail account Columbia already sells through.
What this gives you: the target list. This is where the engines are reading from on these questions, so it is the set worth being present in.
What it does not give you yet: the route into each one. Access differs by publication and none of it is assumable, so mapping it is a research pass of its own before anything is committed.
Community threads — a strategic advocacy program
The buyer data set is where the audit found the sharpest version of every question, and the engines read these threads back — Google AI Overview quotes them by name in the captured answers. A brand account answering in these threads reads as marketing.
The program: equip the people who are already in them — retail staff, fit specialists, ambassadors, warranty and repair — with the facts and the freedom to answer as themselves. Columbia provides the accuracy, and the credibility comes from the person answering.
The risk: these communities recognize a coordinated push. Repeated phrasing across accounts, or someone recommending Columbia without saying they work there, turns the thread into a complaint about the brand.
Doing it right: everyone is transparent about their relationship with the brand, whether employee, retail partner or sponsored advocate. They answer only in the categories they genuinely know, and they are free to say when a competitor is the better product for that person.
It also calls for a different skill set. 5.1 through 5.4 are engineering and merchandising problems with clear owners inside the company. 5.5 is an advocacy program judged by people who recognize a campaign on sight, which is why it belongs with someone who works in these communities rather than at them.
How it gets measured: re-run this prompt set on a schedule. The four attributes and the prompts they came from are the scorecard — a rise in the unbranded 25.5% is the result, not impressions or placements.
All five are engineering, data and positioning work. Publishing more content does not move any of the gaps this audit found.
6Open questions
These are the questions public data cannot answer. Each needs someone inside the company, and each changes what the strategy above would cost.
Is the value position deliberate? The engines have the brand filed as mid-tier value, unanimously. If that is the strategy, “own the trail” means owning the entry tier, and the work is to be the undisputed best value. If it is not, there is a gap between price and positioning, and the review layer will keep siding with the price.
Who owns product data across retailers? Merchandising, e-commerce and channel teams all touch it. One team has to own the drift. At present the answer to “what does this cost” depends on which retailer an assistant reached first.
Do the specs exist internally as structured data? The product pages carry weight in grams and lug height in millimeters. If that is already structured, this is a plumbing project. If it is hand-written per page, it is a different budget entirely.
Has the traffic question been tested? US sales were down 4% last quarter on soft brick-and-mortar traffic and a weaker wholesale environment, against international up 9%. Soft retail traffic has many plausible causes and no outside party can attribute it — but it is testable.
LSX Partners
AI Visibility Assessment & Strategy · September 2026. Prepared as an independent case study. Columbia Sportswear is not a client of LSX Partners and was not contacted. All findings derive from publicly available sources and live AI engine testing conducted by LSX Partners. Figures are point-in-time and carry the date of capture.
Methodology, prompt design and scoring framework © LSX Partners.