Selfridges × Korpora · AI mindshare measurement · luxury department store
Selfridges is a landmark store that AI cannot see.
0 of 54 unprompted luxury answers.
WHO AI CITES ACROSS 12 LUXURY
BUYER-DECISION QUERIES (US)
selfridges.com ··········· null
+7 rivals ··········· null
CITES INSTEAD
youtube ██████████████ 7237
reddit █████████████░ 6635
instagram ██░░░░░░░░░░░░ 1286
wikipedia █░░░░░░░░░░░░░ 315
Invisible unprompted, owns
none of the citations.selfridges.com and seven luxury rivals all return null at the domain level. AI cites YouTube, Reddit and Wikipedia instead. Source: Google AI citation graph, selfridges-v3 run 2026-06-01.
Across 54 subject-blind cells, Selfridges surfaces in 0.0% of AI answers. When a buyer asks an AI assistant where to shop for a luxury occasion and names no brand, a store that anchors Oxford Street simply does not come up. The brand does win when the models are handed its name (88.9% of direct-comparison framings, 100% of differentiation), but that is the model complying with the prompt, not recalling Selfridges on its own. And look at who AI actually cites across luxury buyer-decision queries: selfridges.com returns null, as do all eight rivals. AI leans on third-party authority instead, YouTube, Reddit, Wikipedia. The entire opportunity is building presence on the surfaces the models read; right now there is none to measure.
WHO AI CITES ACROSS 12 LUXURY
BUYER-DECISION QUERIES (US)
selfridges.com ··········· null
+7 rivals ··········· null
CITES INSTEAD
youtube ██████████████ 7237
reddit █████████████░ 6635
instagram ██░░░░░░░░░░░░ 1286
wikipedia █░░░░░░░░░░░░░ 315
Invisible unprompted, owns
none of the citations.selfridges.com and seven luxury rivals all return null at the domain level. AI cites YouTube, Reddit and Wikipedia instead. Source: Google AI citation graph, selfridges-v3 run 2026-06-01.
Selfridges · public dashboard
Selfridges's AI mindshare, live
3 headline views
snapshot · 2026-06-01
Where Selfridges stands in the answers AI shopping assistants give luxury buyers, measured against its cohort: Farfetch, SSENSE, NET-A-PORTER and Mytheresa. Tab through the findings; each number is real, pulled from the v3 run on 2026-06-01.
Action cards
What to ship before the next training corpus closes.
Ordered by qualitative lift potential (cross-stream evidence weight, ship feasibility, and time-to-corpus). The strongest cards give an AI the machine-readable proof to pick you and defend that choice to its operator, in metrics, not adjectives. Each carries a paste-ready engineer prompt; cards involving copy and image generation also carry a brand-voice growth prompt.
Active
5 cards · 0 shipped
In-flight and queued. Every card is an open recommendation today; on a paid engagement they push into the team's tools and ship against this board.
Shipped
0 cards executed
Public build history of the Selfridges engagement. Empty because this is a prospective measurement and nothing has shipped yet. The slot is wired, not decorative.
No cards executed yet
The first executed card appears here the minute it pings back through the protocol above.
How we work
Two steps. You are on step one.
Most AI-visibility vendors sell you fear, then a subscription. Korpora sends you your real number first, free. Selfridges cleared the internal fit review (scale, vertical, AI-channel relevance, execution capacity), so Korpora ran the measurement and shipped this dashboard. There was nothing to request and nothing to buy. What happens next comes down to one thing: whether the measurement in front of you earns the engagement.
- 1
Prospective measurement (you are here)
This is it. Korpora ran the full cross-channel measurement on Selfridges and shipped the result as the dashboard you are reading: the live mindshare monitor, ranked action cards (each with a paste-ready engineer prompt), the full method, and the Ask Korpora chat for follow-up questions. Run once, on us. Nothing was requested and nothing is owed. You read it and decide whether the measurement earns the engagement.
- 2
Paid engagement (what continuing looks like)
When the measurement earns it, continuing is a paid engagement scoped to Selfridges's cohort and depth: weekly re-measurement across every stream and every model, Western and Chinese, fresh action cards as the surfaces move, an alert the moment a rival accelerates into the next training corpus, and an operator who can run the moves alongside your team. There is no fixed self-serve tier; every cohort prices differently. Apply with a work email and we come back with scope and pricing, no call required to get the number.
Ask Korpora
Tell Korpora what you need
Methodology
How we measure, in full.
Every number on this page traces back to one of the streams below. Expand any section for the method, the headline-metric definitions, the live streams themselves, and the caveats.
1 · How we measure
How each number on this page is produced. Every stream re-runs weekly on a paid engagement:
- AI assistant recall: buyer-decision queries (subject-blind and subject-named) run through three frontier models, 108 cells this cycle, Wilson 95% confidence range on every share. Organic (unprompted) reported separately from evaluation (named).
- LLM citation graph: which domains Google AI and ChatGPT cite for luxury department storebuyer queries, and whether any retailer's own domain is cited at all. From DataForSEO LLM Mentions.
- Creative landscape: competitive Meta ad inventory (Facebook + Instagram). Active ad volume per company, from the live GetHookd ad library.
- Instagram + TikTok organic: engagement-per-post and engagement-per-video, content-format mix. Via Apify (live), shown in the attention-footprint stream.
2 · How to read the headline metrics
- Organic owned share (0.0%, the metric that drives action)
- Selfridges was surfaced by AI without being named in the prompt. We asked the model an open buyer-decision question and recorded whether Selfridges came up. 54 subject-blind cells; Selfridges was named in 0 of them. This is unprompted recall: whether AI surfaces Selfridges to a buyer who does not already know it exists. We treat it as a leading proxy for AI-driven discovery, not a measured conversion.
- 94.4pp organic-to-evaluation spread = cold start classification
- The gap between organic (0.0%) and evaluation (94%) is 94.4 percentage points. A spread of 15 to 40pp is what Korpora flags as a moderate gap (our own working threshold, not an industry standard; above 40pp is cold-start): the model knows Selfridges well enough to engage when named but does not surface it unprompted at the discovery stage as often as you would want for a category-defining company.
3 · The live measurement streams
Three data streams, re-run weekly on a paid engagement. Streams 1 and 2 (AI recall, citation graph) are the model-memory readout; Stream 3 collapses the attention channels (Meta ads, Instagram, TikTok), none of which feed the training corpus, into one panel that sizes the attention-vs-memory gap.
Stream 1 of 3 · live data · Korpora 108-cell battery · 2026-06-01
AI assistant recall composition
12 buyer-decision queries × 3 frontier models (Claude Sonnet, Claude Haiku, GPT-5.5) × 3 rounds, 108 cells total. Subject-blind cells (organic) reported separately from subject-named (evaluation).
Organic mindshare composition (54 subject-blind cells)
100% of buyer-decision queries are either contested or unclaimed, the mindshare we would treat as addressable over the next two corpus cycles.
Subject-own organic AI mindshare · head-to-head (each brand’s own battery)
On each brand's own subject-blind battery, Selfridges lands at 0.0% (0/54 cells, Wilson 95% CI 0.0-6.6%), tied with Bergdorf Goodman at the bottom of the cohort. Farfetch leads at 61.1% and SSENSE, measured in its own reference run, sits level at 61.1%, so the co-leaders set a 61.1% ceiling; NET-A-PORTER trails at 44.4% and the rest collapse from Mytheresa's 9.3% down to zero. The honest read is that Selfridges is not in the conversation when a buyer names no brand. The CI ceiling of 6.6% is the math being conservative on a zero count, not a hidden signal; across three independent query rounds the subject share is 47.2% ± 4.8pp on aggregate, but every one of those wins comes from framings that name Selfridges in the prompt. Strip the name out and the count is zero, three rounds running.
Cross-battery recall
- The organic-to-evaluation spread is the widest in the cohort: 0.0% unprompted against 94.4% when the brand is named in the query (51/54 subject-named cells).
- That 94.4pp spread classifies as cold start: the evaluation wins are the models complying with a named brand or hallucinating capabilities, not a trained association they reach for on their own.
- By framing, the split is absolute: direct-comparison 88.9% (24/27) and differentiation 100.0% (27/27) when the brand is named, but use-case 0.0% (0/27) and discovery 0.0% (0/27) when it is not.
This is the cleanest cold-start signal in the cohort. The models can run a "Selfridges vs Mytheresa" comparison on demand and will default to Selfridges as the broader-range, UK-convenient anchor, but they cannot volunteer the brand when a buyer opens with a problem and no name. Treat the 94.4% evaluation share as query-framing artifact, not buyer signal: it measures the models' willingness to answer a leading question, not what they know. The entire job is moving unprompted recall off zero before the next training cycle indexes.
Across both views, Selfridges is a cold-start in luxury AI discovery: zero unprompted recall against co-leaders sitting at 61.1%, and an evaluation-stage number that evaporates the moment you stop naming the brand. The composition makes the path concrete, of the 54 subject-blind cells, 0 are owned, 43 are contested (a rival surfaced and Selfridges did not) and 11 are unclaimed (no cohort brand surfaced). The work is not defending a position; it is creating one, on the surfaces models read (Stream 2), where Selfridges, like every rival, currently owns none of the domains AI cites.
Stream 2 of 3 · live data · DataForSEO LLM Mentions · 2026-06-01
LLM citation graph
For 12 luxury buyer-decision queries (US-scoped), which domains do Google AI and ChatGPT cite most as sources?
Google AI · top cited domains
ChatGPT · top cited domains
Domain-scoped LLM Mentions returns null for selfridges.com on both platforms. Same for farfetch.com, ssense.com, mytheresa.com, net-a-porter.com, mrporter.com, highsnobiety.com and bergdorfgoodman.com. No luxury retailer is cited at the domain level. AI leans entirely on third-party surfaces, video, forums, resale marketplaces and reference sites, when it answers a luxury buyer. For a cold-start brand this null is double-edged: Selfridges is invisible like everyone else, but it also has no organic recall to fall back on, so the citation surfaces are the only lever, not a supplement to an existing lead.
All LLM Mentions counts are US-scoped (DataForSEO location 2840), Google AI and ChatGPT pulled as separate platforms, summed across 12 buyer-decision queries. The cohort pattern (no luxury retailer at domain level) is robust; individual per-domain ordering on a single query is less reliable. DFS billed about $0.10 per query.
Stream 3 of 3 · attention footprint vs AI memory
Meta ads, Instagram, and TikTok are channels models do not read. None feed the training corpus, so none move organic AI mindshare directly. They are measured together here to size one gap: the attention Selfridges buys and earns on human channels, against how little of it compounds into the model memory measured in Streams 1 and 2.
Paid signal · Meta ad inventory · GetHookd
Active Meta ad inventory
Count of unique active Meta ads (Facebook + Instagram placements combined) per brand, from the GetHookd ad library.
Selfridges runs 28 active Meta ads, a mid-pack presence well behind Mytheresa (602) and NET-A-PORTER (211). But the inversion is the point: Farfetch and SSENSE co-lead organic AI recall at 61.1% while running 0 and 98 Meta ads respectively, and Selfridges, at 0.0% organic, is spending into the feed with nothing to show for it in the channel that matters here. Paid Meta volume does not buy AI mindshare in this cohort. The 28 ads keep Selfridges visible to shoppers scrolling Instagram and Facebook, but feed impressions are not what the models read, so this spend is not the route off a cold start.
Data gaps: Farfetch (no active Meta ads in the library).
Snapshot from the GetHookd ad library, 2026-06-01. Active-ad counts combine Facebook and Instagram placements. Farfetch returns no tracked active ads; for brands of their size this most likely reflects a deliberately low Meta presence rather than a data gap.
Organic signal · Instagram
Instagram organic engagement-per-post
Mean of (likes + comments) on the most recent 30 posts per brand. Via Apify.
Selfridges is absent from this chart: the Instagram pull resolved to a wrong handle and was excluded rather than reported as a false number. So there is no valid Instagram engagement datapoint for Selfridges in this run, and the cohort ranking above is read without it. What the chart does show is the surface Selfridges would be competing on, editorial-led accounts like Highsnobiety (11,728) and SSENSE (3,313) set the pace, and Instagram engagement, in any case, is not what moves AI recall. The TikTok read below is the social signal we can actually anchor for Selfridges.
Most-recent posts per brand, via Apify, captured 2026-06-01. Bootstrap roughly ±15% on engagement averages at this sample size. The Selfridges Instagram pull resolved to a wrong handle and is excluded, so Selfridges has no valid IG datapoint this run; treat the gap as missing data, not a zero. MR PORTER's @mrporterlive is likely a secondary account.
Organic signal · TikTok
TikTok organic engagement-per-video
Mean of (likes + comments + shares + saves) on the most recent 30 videos per brand. Via Apify.
Selfridges has the third-largest TikTok following in the cohort (179.4K) but one of the lowest engagement rates measured: 150 average engagements per video, a 0.08% ER, ahead only of SSENSE and Farfetch on raw engagement and tied with Farfetch at the floor on rate. The follower count is a real-world brand asset that has not translated into the kind of high-engagement, repeatedly-clipped content that seeds the corpus. StockX (32,810) and Vestiaire Collective (7,388) show what an engaged resale audience produces; MR PORTER drives 3,240 per video off a fraction of Selfridges' followers. As elsewhere, this surface is not where AI recall is built, but it is the clearest evidence that audience size and indexable attention are not the same thing for Selfridges.
Data gaps: Mytheresa, Bergdorf Goodman (squatted or unverified handles).
Most-recent videos per brand, via Apify, captured 2026-06-01. Engagement is volatile post-to-post. The Selfridges TikTok datapoint (@selfridges, 179.4K followers, 0.08% ER) is valid; the Instagram pull for the same brand was not and is excluded above. @mytheresa and @bergdorfgoodman resolve to squatted or empty handles and are excluded.
4 · Notes & caveats
- Query battery. This run uses buyer-decision queries inferred from the category, not Selfridges's real on-site search terms. A paid engagement swaps the inferred queries for the terms buyers actually use, so the organic-mindshare number and the per-query gaps get materially more representative.
- Card lift quantification. Cards give the mechanism, the hypothesis, and the re-measurement criterion, not yet a calibrated forecast. The ordering is a qualitative lift ranking.
- SSENSE's organic figure (61.1%) is from its own reference run (2026-05-28); the rest of the cohort is from the selfridges-v3 run (2026-06-01). Same battery shape, subject swapped, so the comparison stays apples-to-apples.
- Selfridges' aggregate share (47.2%) looks mid-pack, but it mixes subject-named and subject-blind framings and is reference-only; the 94.4pp cold-start spread is why organic (0.0%), not aggregate, is the headline.
- The Selfridges Instagram engagement pull resolved to a wrong handle and is excluded, so there is no valid IG datapoint for Selfridges this run; the TikTok datapoint (@selfridges) is valid.
Re-measurement cadence: weekly across every stream. The Korpora measurement model and the channel surfaces it measures both evolve continuously; weekly captures inflections that quarterly would miss.
Methodology question or factual correction? Ask in the chat above.
Related: full methodology · data provenance · luxury leaderboard · home