1,125 responses, 5 models, 45 Turkish questions: the baseline AI-visibility snapshot of Turkey's home textiles market

Vertical: Home textiles (ev tekstili), Turkey
Method: A single point-in-time study of 1,125 responses across five large language models (Claude, Gemini, GPT, Perplexity, Grok), each asked 45 Turkish-language questions five times
Collection window: 8 August 2026

Key terms: AI visibility, AI Visibility Score, home textiles brands, ev tekstili, nevresim takımı, havlu, perde, Turkish (yerli) home textile brands, LLM brand recommendations, ChatGPT/Gemini/Claude/Perplexity/Grok brand recommendation, Generative Engine Optimization (GEO).

This report reproduces the real Turkish-language questions people ask AI assistants about bedding, towels, curtains and household textiles, and measures how often, in what order, and across how many models five large language models name each brand. Because the same questions can be read with or without an origin restriction, and because home textiles divides into genuinely different product categories, the entities are ranked in ten separate cuts, defined in Chapter 2. It is a single point-in-time study and should be read within the limitations in Chapter 11.

How to read this report: visibility is not quality

Every number here measures whether an AI assistant names a brand, and how early. It does not measure whether the towels are absorbent, whether the sheets last, whether the price is fair, or whether the recommendation was accurate. A brand can be highly visible and mediocre, or excellent and invisible. Read the leaderboards as a map of findability in a new discovery channel — nothing more.

1. Executive Summary

One-sentence takeaway: Taç leads Turkey’s home-textile shelf across almost every cut, and the domestic industry leads whether or not the question asks for Turkish brands, leaving the foreign presence confined to retailers and a licence.

• Taç leads decisively. It tops the general leaderboard at 96.59, appearing in 56.18% of all 1,125 responses — 632 answers — and named by all five models. Özdilek (85.91, 48.80%) and Karaca (80.80, 42.13%) follow, with Yataş (77.06) and Chakra (68.74) completing the top five.
• 20 entities qualified for the general ranking. Of 266 tracked entities, 20 cleared the 5% mention threshold across all 45 questions: 18 Turkish and 2 foreign. Their average score is 59.1, and 19 of the 20 are named by all five models.
• Domestic visibility here is unconditional. On the 23 origin-neutral questions, Turkish entities take 89.2% of mentions and 18 of 22 qualifying places. On the 22 questions that explicitly ask for yerli brands, that rises to 98.7% and all 22 qualifiers are Turkish. The origin clause changes the answer remarkably little, because the origin-neutral answer is already domestic.
• The foreign presence is retail, not brand. The only foreign entities to qualify in the open cut are IKEA (#7), the licensed Pierre Cardin (#17), Zara Home (#19) and H&M Home (#21) — a furniture retailer, a licence, and two fashion retailers’ homeware lines. Across all ten cuts, exactly one foreign home-textile brand qualifies anywhere: Brooklinen, which clears the threshold in Towels at precisely 5.00% on 10 responses from a single model, and finishes last of that segment’s 28 qualifiers.
• Four segments, four leaders. Yataş leads Bedding & Quilts at 63.78% of responses; Chakra leads Towels at 67.00%; Taç leads Curtains at 77.71%, the highest mention rate recorded anywhere in this study; and Özdilek leads the Multi-Category cut at 59.33%. Taç is the only entity in the top four of every segment.
• The models differ more in how they answer than in whom they name. Grok names 10.43 entities per answer and Gemini 9.83, against 4.52 for GPT-5.6 and 4.60 for Claude Haiku. Taç is the most-named entity in three of the five models; Perplexity puts Özdilek first, and Grok answers with marketplaces — Trendyol in 97.8% of its answers and Hepsiburada in 92.0%, ahead of any brand.
• One model barely searched. Overall 86.0% of answers were web-grounded, but that average conceals a split: Claude, Perplexity and Grok grounded 100% of their answers and GPT-5.6 99.6%, while Gemini grounded only 30.7%. 155 of the 157 ungrounded responses in the entire study are Gemini’s — which makes this the first edition where a search-versus-memory comparison is possible at all (Chapter 6).
• Sources are brand-led. Brand-owned sites account for 52.82% of citations — more than half — against 15.61% “other”, 10.91% editorial, 10.36% marketplace and 5.32% forum/social. Seven of the twelve most-cited domains belong directly to entities in the leaderboard.

Why it matters. Shoppers are shifting discovery from search engines and marketplaces to AI assistants. Independent survey work suggests roughly four in ten Turkish internet users were using ChatGPT monthly in mid-2025 (GWI), while TÜİK’s first measurement of the subject put self-reported generative-AI use at 19.2% of people aged 16–74 in 2025 — different definitions, both indicating a channel that now exists at scale. Home textiles is a category where Turkey has genuine industrial depth, and this edition asks a narrow question about it: when someone asks an assistant in Turkish which sheets or towels to buy, whose name comes back. This is the first (baseline) snapshot of that shelf. One reminder: the numbers describe visibility and information availability, not which brands are best.

2. Methodology

One-sentence takeaway: Five large language models were asked 45 Turkish questions five times each, with no system prompt, producing 1,125 responses. Entities were identified by alias matching plus human review and ranked with a three-component (45/30/25) score, computed separately within each of ten cuts.

2.1 Scope and models

The study queried five large language models directly via API with identical questions. All models received only the user question. No system prompt was used, temperature was not set manually (provider default), and no region or locale parameter was defined. Because the questions are in Turkish, the models infer the Turkish market from language alone.

Scale: 45 questions × 5 repeats × 5 models = 1,125 responses. Every cell of the grid existed exactly once, with no duplicates, and every response returned a completed status with a non-empty body — so collected equals analysed, and no denominator required adjustment. Of the 1,125, 1,118 named at least one tracked entity; seven named none. Single collection window, 8 August 2026, 11:54–15:38 UTC.

The final entity dictionary carries 266 entities across 402 aliases, of which 232 are Turkish, 32 foreign and 2 of unknown origin — an overwhelmingly domestic dictionary, which is worth holding in mind against the leaderboards that follow.

The ten cuts. Every metric is computed within a cut, against that cut’s own denominator. Rates from different cuts are not comparable, and neither are scores.

2.2 Web-search configuration

Web search was offered as an enabled tool to all five models; each decided for itself whether to use it (Perplexity’s Sonar is search-grounded by design). Whether a response drew on web search or the model’s own knowledge was inferred from whether the provider returned any citation or source: a citation means “web-search,” none means “own knowledge.” This is a reasonable proxy, not direct proof of the model’s internal process.

In this run 86.0% of answers were grounded, but the average is misleading. Four of the five models grounded on essentially every answer; Gemini grounded on 30.7% of its own. Any model-level comparison in this report has to be read with that asymmetry in view, and Chapter 6 treats it as a finding rather than a footnote.

2.3 Questions and segments

The 45 questions reproduce the real Turkish-language queries people bring to an assistant about home textiles: general discovery (“ev tekstilinde hangi markalar gerçekten iyi”), specific attributes (cotton quality, thread count, absorbency, dye fastness, domestic production), and use cases (dowry sets, baby rooms, summer bedding, blackout curtains, hotel-quality towels). They are balanced globally across the three behavioural types — 15 discovery, 15 attribute, 15 use_case — and split 22 local against 23 open. No question rewriting was required for this edition.

The seeded subcategory field was empty throughout, so segment leaderboards are built on a four-way product rollup:

Of the 45, 22 are “local” (they explicitly ask for Turkish / yerli brands) and 23 are “open” (no origin restriction). The Turkish-versus-foreign comparison in Chapter 8 rests solely on the 23 open questions, the only fair basis for comparing origins, since the local questions cannot surface foreign brands by construction.

2.4 Entity extraction and classification

Entities were extracted by alias-based matching, word-boundary anchored, with Turkish-aware case folding — İ and ı are normalised explicitly, since naive lowercasing corrupts them. Candidates were filtered by a data-driven capitalisation ratio with a domain-citation override, so lowercase-styled or domain-backed brands were not lost. 982 candidates were auto-dropped with a stated reason each; 653 went to human review, resolving to the final 266 entities and 402 aliases.

A stricter brand-visibility rule. A product-line alias credits its parent brand only in responses where the parent is independently identifiable. Where a model named a product line without naming the brand that owns it, the brand was not credited — this study measures brand visibility, and a reader of that response could not have recovered the brand from it. These cases are counted separately: 11 orphan observations across 5 entities. The clearest example is IKEA, which was named through product names such as Hälsa in three responses that never mentioned IKEA itself.

Origin follows the trademark, not the operator. A brand is Turkish only if the mark itself is Turkish-owned, derived from the trademark registration against Türkiye. Foreign marks operated in Turkey under Turkish licence are classified Foreign and tagged Licensed-brand so they can be isolated. Two entities carry an unknown origin (Noémie and Thermoclean) and are excluded from every origin share.

Entity types. Six classes are tracked: Brand, Brand+Retailer, Retailer, Marketplace, Licensed-brand and Non-brand. Brand+Retailer is the characteristic type of this vertical and covers 79 of the 266 entities, including almost every leader — Taç, Özdilek, Karaca, Yataş, Chakra, English Home and Linens all design, manufacture and operate their own store networks. Separating “brand” from “retailer” would misdescribe them.

What the brand-only views remove. Only the eight marketplaces: Akakçe, Amazon, Eperde, Etsy, Hepsiburada, N11, Perdenette and Trendyol. Vertically integrated own-label retailers such as IKEA, Koçtaş, JYSK, Boyner and Beymen are retailers rather than marketplaces and are retained in every table. In a vertical where the leading brands are retailers, removing retailers would empty the leaderboard.

Scope fit. 173 entities are classed Core, 63 Adjacent and 30 Out-of-scope. Adjacent entities — furniture, homeware and fashion brands with genuine home-textile ranges, including Karaca and IKEA — are included in all cuts, because they answer questions the study deliberately asked. The 30 out-of-scope entities appear in no leaderboard and no origin share, and fall into three groups: 20 manufacturers and corporate parents typed Non-brand which models named instead of a consumer brand, 7 ingredient trademarks (Woolmark, Dacron, Hollofil, Allerban, Thermolite, Evolon, Thermoclean), and 3 adjacent-market brands. Chapter 11 explains why this matters more here than usual.

Definitions.

• Mention: a validated entity (or alias) appears in the answer text, counted once per response (presence, not frequency).
• Position (MRR): mean reciprocal of the entity’s first-appearance rank within each answer (first = 1.0; second = 0.5; …), averaged only over the responses where it appears.
• Breadth: number of the 5 models (0–5) that mention the entity within that cut.

2.5 AI Visibility Score (0–100)

Score = 0.45 × Mention + 0.30 × Position + 0.25 × Breadth

This edition changes how the score is scaled, and the change matters for anyone comparing editions. Each component is divided by the highest value observed among qualified, in-scope entities within that cut, so the cut’s leader on any one component scores 100 on that component. A score of 100 overall therefore requires a single entity to top all three components at once — which happens four times in this study (Taç in the open, Curtains and use_case cuts, Yataş in Bedding & Quilts), and does not happen in the general cut, where Taç scores 96.59 because Hamam, not Taç, tops the position component.

Two consequences follow, and both are load-bearing:

1. Scores are not comparable across cuts, editions or verticals. The scale moves whenever the leader moves, so 100 in one cut is a different absolute quantity from 100 in another. For any comparison beyond this edition, use the raw layer — mention rate, MRR and breadth — which is published for every entity in Chapter 12.
2. An unqualified entity can exceed 100 on a component. Unqualified entities are scored against a scale they did not help set, so 77 of the 1,205 unqualified in-scope rows (6.39%) carry at least one component above 100, the highest being a position component of 183.75. These are flagged and never clipped. No qualified entity can exceed 100 by construction, and none does.

Qualification (≥5% rule): only entities mentioned in at least 5% of that cut’s responses are ranked. This limits the risk that a rare brand which happens to appear first in a few answers inflates the position component. It does not eliminate it — see §3.1 and Chapter 11. Unfiltered metrics for all 266 entities appear in Chapter 12.

2.6 Neutrality and self-exclusion

This is a market-wide, neutral study with no focus brand. To prevent any conflict of interest, all citations to Herm.io’s own domains were excluded from the source data before analysis, so the company’s own content neither appears among the most-cited domains nor influences any figure.

3. Overall Visibility Leaderboards

One-sentence takeaway: Taç leads the general, local and open cuts alike, and the composition of all three is overwhelmingly domestic — the three headline cuts tell the same story.

As a reminder, these rank visibility, not product quality, durability, price or reliability.

3.1 General leaderboard: 20 qualified entities

Average score of the 20 qualified entities: 59.1. Origin split: 18 Turkish, 2 foreign.

Reading. Taç tops the table at 96.59, named in 56.18% of answers — 632 of 1,125 — by all five models, and it leads on volume by a wide margin: Özdilek at 48.80% is the only other entity above 45%. The top six are all Turkish Brand+Retailer entities, businesses that manufacture and retail under their own name. 19 of the 20 qualified entities are seen by all five models; the exception is Hepsiburada at #18, on four.

The two foreign entries are instructive for what they are not. IKEA at #9 is a furniture retailer whose textiles ride along with its catalogue, and Pierre Cardin at #20 is a licensed mark rather than a foreign home-textile business operating in Turkey. There is no foreign equivalent of Taç in this data.

The entry at #7 requires explanation. Hamam ranks seventh on a 5.60% mention rate — 63 responses, against Özdilek’s 549, an entity named nearly nine times more often. Its MRR of 0.617 is the highest in the cut, meaning that on the rare occasions it is named, it is named early. Because the position component is scaled against that same maximum, Hamam scores 100 on position, and the 30% weight is enough to carry it past entities with far more volume. Hamam is a real and well-established Turkish towel and bathrobe brand — but its rank here reflects the score’s design as much as its share of the conversation. The next table makes the effect visible rather than asking readers to take it on trust.

3.2 The same cut, ranked by mention rate

Identical data, ranked by volume alone rather than by the weighted score:

The reordering is instructive. Hamam leaves the top ten entirely. Hepsiburada climbs from #18 to #10 and Trendyol from #12 to #9, because both are named often but late — their MRRs are 0.151 and 0.225 against Taç’s 0.547. Madame Coco rises from #10 to #6 on the same logic. Moving the other way, Chakra falls from #5 to #7 and IKEA from #9 to #11, because both are named comparatively early when they appear.

Neither ranking is more correct than the other: the first asks how prominently is this brand named, the second asks how often. The AI Visibility Score deliberately blends the two, and readers should hold both pictures in mind for the rest of the report.

3.3 Brand-only view (marketplaces removed)

Two of the 20 qualified entities are marketplaces rather than brands: Trendyol (#12) and Hepsiburada (#18). They are kept in the headline ranking because that is how shoppers meet the shelf, but removing them gives a pure brand-versus-brand picture:

Neither marketplace sits high enough to distort the top of the table: removing them leaves the leading eleven otherwise unchanged, with Taç, Özdilek, Karaca, Yataş and Chakra still first through fifth. Retailers such as IKEA and the Brand+Retailer entities stay in, and §2.4 explains why. In a vertical where the leading brands operate their own stores, a ranking that stripped retailers would remove the market itself.

3.4 Local leaderboard: 22 qualified entities (Turkish-only questions)

When the 22 questions that explicitly ask for Turkish/yerli brands are isolated, the shelf does not invert — it intensifies:

Average score of the 22 qualified local entities: 53.9. Origin split: 22 Turkish, 0 foreign.

Reading. Taç leads at 92.62 on 48.18% of local-cut responses, with Özdilek close behind at 89.83 — a much narrower gap than in the general cut. Karaca (80.00), Yataş (73.31) and Chakra (70.50) follow in the same order as the general leaderboard. Not a single foreign entity clears the threshold, and Turkish entities take 98.7% of the cut’s mentions.

That is a stricter result than it may appear. Long-established foreign names that Turkish consumers might reasonably think of as local do not slip through: the filter is respected absolutely, which suggests the models hold a clear and correct picture of which home-textile brands are Turkish.

The lower half of this cut is also where the data thins. Cottonin Home (#21) is named by only two models and Doqu Home (#22) by three, against five for most of the top ten. Breadth is the component doing the work at the bottom of this table, and small differences there should not be over-read.

4. Segment Leaderboards

One-sentence takeaway: There is no single home-textiles shelf. Four product segments produce four different leaders, and only Taç places in the top four of all of them.

The four segments partition the same 45 questions by product category. Because question types are not balanced within segments — Multi-Category carries no attribute questions at all — counts of distinct entities should not be compared across segments; see Chapter 11. Rankings within each segment are unaffected.

4.1 Bedding & Quilts (18 questions, 450 responses)

The largest segment, and the only cut Taç does not lead. Yataş takes it at 100.00 on a 63.78% mention rate — it tops all three components — with Karaca (93.97, 59.56%) second and Taç third. Yataş and Karaca are both classed Adjacent: a mattress and sleep-systems business and a homeware group respectively, both with substantial bedding ranges. When the question is about duvet sets and quilts, the assistants reach past the pure textile houses toward the sleep and homeware brands. 24 of the 27 qualifiers are Turkish.

Penelope at #5 (65.76 on 19.33%) is the strongest position-driven entry here, and Ecocotton at #7 qualifies on 7.33% with an MRR of 0.527 — an organic-cotton specialist named early when named at all.

4.2 Towels (8 questions, 200 responses)

The most concentrated segment at the top and the most domestic overall: 27 of its 28 qualifiers are Turkish, the highest domestic share of any cut in the study. Chakra leads at 94.98 on a 67.00% mention rate, with Özdilek almost level at 91.28 (66.50%) — the two are separated by position, not volume. Taç falls to #4 here despite matching English Home’s 40.50% mention rate, because it is named noticeably later in towel answers (MRR 0.265 against English Home’s 0.313).

Hamam reappears at #6 on 18.00%, a far more solid footing than its general-cut placing, which is consistent with it being a genuine towel specialist rather than a scoring artifact. The segment’s single foreign qualifier is Brooklinen, last at #28 on exactly 5.00% from one model — the thinnest entry in the cut, and thinner than DK Store at #11 (7.00% on two models).

4.3 Curtains (7 questions, 175 responses)

The smallest segment, and the most dominated. Taç reaches 77.71% of responses with an MRR of 0.811 — the highest mention rate and the highest position score anywhere in this study — and takes a perfect 100.00. Brillant follows at 79.19 on 65.14%. After those two the segment fragments quickly into small specialists: İ.D İpek Dokuma, Persan, Perva, Perdeland, Özenev, Dali Tekstil, Fullplise.

Curtains is also where the data is weakest, and the fragmentation is the reason. Nine of its 30 qualifiers are named by fewer than three models, and all three of the study’s sub-10-response qualifiers sit here — NK Toprak Stor, Perden Tekstil and Vesna Curtain Fabric, each on exactly 9 responses. On 7 questions and 175 responses, the 5% threshold is met by 8.75 responses. The leaderboard is published in full because the pattern below the top two is itself a finding: curtain visibility is genuinely model-specific in a way no other segment is. But entries below roughly #7 should be read as indicative, not settled.

4.4 Multi-Category (12 questions, 300 responses)

The whole-home cut, where questions ask about home textiles in general rather than a product. Özdilek leads at 88.19 on 59.33%, narrowly ahead of Taç (87.43, 57.67%) — the only cut Özdilek wins. It also produces the most qualifiers of any segment (36), because general questions surface long lists.

This is the cut with no attribute questions at all (6 discovery, 6 use_case), so it reflects discovery and use-case phrasing only, and its distinct-entity count should not be compared with other segments.

4.5 Segment comparison

Three observations. First, no entity leads more than one segment — the four leaders are four different companies. Second, Taç is nonetheless the only entity to place in the top four of every segment, which is what carries it to the top of the general cut without leading the largest one. Third, the foreign share is thin everywhere and thinnest in towels, where a single foreign entity qualifies out of 28.

5. Differences Between Models

One-sentence takeaway: The models broadly agree on Taç and Özdilek, but they differ enormously in how many names they give — Grok lists more than twice as many entities per answer as GPT-5.6 — and Grok alone answers with marketplaces.

5.1 Per-model behaviour summary

The clearest split is breadth of answer. Grok names 10.43 entities per answer and Gemini 9.83, while GPT-5.6 (4.52) and Claude Haiku (4.60) name less than half as many; Perplexity sits between at 6.40. But breadth per answer and breadth of vocabulary come apart here in a way worth noting: Gemini names the second-most entities per answer yet has the narrowest vocabulary in the study at 111 distinct entities, against Perplexity’s 160 and Claude’s 149. Gemini gives long answers drawn from a short list. Perplexity gives medium answers drawn from a long one.

This matters for interpretation. A brand in the long tail has a materially different chance of being named depending on which assistant the shopper opens. In Grok or Perplexity, an answer has room for specialists; in GPT-5.6, an answer listing four or five names rarely goes beyond the best-known houses.

5.2 Each model’s most-named entities

The tables below give each model’s five most-mentioned in-scope entities (rate = % of that model’s 225 answers naming the entity).

Taç is the most-named in-scope entity in three of the five models — Gemini (89.8%), GPT-5.6 (54.2%) and, jointly with Karaca, Claude Haiku (22.2%). Perplexity puts Özdilek first at 45.8%.

Grok is the exception, and instructively so. It names Trendyol in 97.8% of its answers and Hepsiburada in 92.0%, putting both ahead of every brand; Karaca, its highest-ranked brand, reaches 72.9%. Grok answers Turkish home-textile questions substantially by pointing at where to buy rather than what to buy — a behaviour no other model in the set shares at anything like that rate.

The intensity differences are as large as the ordering ones. Gemini names Taç in 89.8% of its answers; Claude names its own top entity in 22.2%. Two assistants asked identical questions produce answers that differ less in who appears than in how consistently anyone does.

A note on what these tables exclude. Each model also named entities that appear in no leaderboard — manufacturers, corporate parents and ingredient trademarks classed out-of-scope. The pattern is not evenly spread. Gemini named Zorlu Holding in 37.3% of its answers and Zorlu Tekstil in 20.0%; Claude named SANKO Havlu in 10.7% and Ağaoğlu Tekstil in 7.6%. These are the parent companies and manufacturers behind consumer brands — Zorlu is the group behind Taç — and a shopper asking which sheets to buy is not helped by the name of the holding company. It is a real difference in answer quality that the leaderboards, by design, do not show.

6. Search vs. Memory

One-sentence takeaway: One model barely searched, and comparing its grounded answers against its ungrounded ones shows that grounding widened its repertoire slightly without changing who led — but the comparison is a single-model finding, not a general one.

The aggregate grounding rate of 86.0% conceals a near-binary split between models.

155 of the 157 ungrounded responses in the entire study are Gemini’s. That fact governs how this chapter can be read. An all-models comparison of grounded against ungrounded answers would not measure grounding at all — it would compare Gemini against the other four models and call the difference a search effect. The only defensible comparison is within Gemini.

Reading. Grounding changes the shape of Gemini’s answer more than its content. When it searches, it names slightly more entities per answer (10.35 against 9.60) and draws on a wider set (77 distinct entities against 66) — but the leader does not move. Taç heads both lists, at 91.3% grounded and 89.1% ungrounded, and the brands beneath it are largely the same names in a similar order.

Two readings are available and the data does not decisively separate them. Either Gemini’s stored knowledge of Turkish home textiles closely matches what a live search returns — which would be unsurprising in a market this concentrated, where the same handful of names dominate both the models’ training data and the current web — or its grounded answers are drawing on sources that mostly confirm what it already held. Either way, the practical implication is the same and worth stating plainly: for this vertical, a brand cannot rely on strong search visibility alone to reach Gemini, because two-thirds of that model’s answers never look.

The other four models ground on essentially every answer, so this report largely measures what they retrieve and select, not what they know unprompted — and retrieval is sensitive to the live web, which changes independently of the models. The grounding rate is measured every edition; if Gemini’s rises, this comparison will be worth repeating on firmer ground.

7. Question-Type Ownership

One-sentence takeaway: The same four names lead all three question types, and the only thing that really changes is how many entities get a mention at all.

The 45 questions are read through three behavioural types: discovery (general “best / recommended”), attribute (specific qualities such as cotton quality, absorbency, thread count or domestic production) and use_case (situation-driven, such as dowry sets, baby rooms, blackout curtains or summer bedding). Each is a cut in its own right with an identical denominator of 375 responses.

Reading. The stability is the finding. Taç leads all three types, and Özdilek, Karaca and Yataş occupy the places behind it in every one, in almost the same order each time.

What does change is the width of the answer. 30 entities clear the 5% threshold on discovery questions, 24 on attribute and 23 on use_case. Open “recommend me a home textiles brand” questions surface a longer list; specific questions concentrate on fewer names. The same pattern appears in the leaders’ rates: Taç is named in 66.13% of discovery answers but 48.53% of use-case ones, because use-case answers spend their length on the situation rather than on listing brands.

Two entities are worth pulling out of the specific cuts. İ.D İpek Dokuma places #6 on attribute questions with a 6.40% mention rate and an MRR of 0.656 — a silk-weaving specialist that the models reach for when asked about fabric quality, and almost never otherwise. Ecocotton places #9 on use_case at 8.53%, with organic-cotton and baby-room questions carrying it. Both are examples of a real pattern in this data: narrow specialists are invisible in general questions and prominent in the specific ones they answer.

The caveat from Chapter 2 applies here. Because question types are not balanced within segments — Multi-Category has no attribute questions, and Bedding & Quilts carries 8 of the 15 — part of any type-level pattern is design rather than market.

8. Open Market: Turkish vs. Foreign

One-sentence takeaway: Across origin-neutral questions, 89.2% of mentions go to Turkish brands, and adding the yerli filter changes remarkably little.

This chapter relies only on the 23 open questions with no origin restriction (575 responses), the only fair basis for comparing Turkish and foreign brands. A blended figure across all 45 questions would be misleading, because 22 of them constrain the answer to domestic brands by construction.

8.1 Origin share

Origin share can be measured two ways, and both are reported here because they answer different questions:

The two measures diverge, and the direction matters. Turkish brands hold 81.8% of the positions on the shelf but 89.2% of the volume, because the largest entities are Turkish: Taç alone accounts for 367 open-cut mentions, more than five times the combined total of all four qualifying foreign entities.

The local cut barely moves it. On the 22 questions asking for Turkish brands, Turkish entities take 98.7% of mentions and 100% of qualifying places. The shift from 89.2% to 98.7% is real but modest — 9.5 points. The origin clause changes the answer far less than it might, because the origin-neutral answer was already domestic.

8.2 Open-market top 15 entities (any origin)

8.3 The pattern: a domestically-owned shelf

The individual ranking shows what the aggregate hides. The open-market top six are entirely Turkish — Taç, Özdilek, Karaca, Yataş, English Home, Chakra — and the first foreign name is IKEA at #7, on a 28.00% mention rate. Below that the foreign entries thin out fast: Pierre Cardin (#17, a licence), Zara Home (#19) and H&M Home (#21).

The composition of that foreign set is the real finding. Not one of the four is a foreign home-textile brand. IKEA and Zara Home are retailers whose textiles are part of a broader catalogue; H&M Home is a fashion retailer’s homeware line; Pierre Cardin is a French mark operated in Turkey under licence, and by this study’s origin rule it counts as foreign even though the business naming it is domestic.

Across all ten cuts and 266 tracked entities, exactly one foreign brand whose primary business is home textiles clears the 5% threshold anywhere: Brooklinen, the American bedding and bath company, which qualifies in the Towels segment at precisely 5.00% — 10 of 200 responses, all from a single model — and places last of that segment’s 28 qualifiers. It appears in no other cut. That is the whole of the foreign home-textile presence in this data, and it rests on one model’s answers.

This is a description of model behaviour, not a claim about the industry — but here the two are unusually easy to reconcile. Turkey is one of the world’s largest home-textile exporters, with deep manufacturing clusters in Denizli, Bursa and Uşak, and several of the entities in this leaderboard are vertically integrated manufacturers with their own retail networks. An AI shelf led by Taç, Özdilek, Karaca and Yataş is not a surprising picture of that market. It is roughly the picture a well-informed Turkish shopper would draw.

This is what a domestic industry that is both strong and findable looks like on an AI shelf. The interesting question for future editions is therefore not whether Turkish home-textile brands are visible — they clearly are — but which of them, and whether the concentration at the top loosens.

9. The Discovery Ecosystem: Where AI Learns About Brands

One-sentence takeaway: More than half of all citations come from brands’ own websites — the assistants are largely reading what the brands themselves publish.

When an assistant answers with web search, it draws on the content available to it at that moment. We collected the source addresses returned by the models, reduced them to domains, and ranked them by how many distinct answers cited each, giving a map of where these systems read about home textiles. Across the run the models cited 528 distinct domains.

9.1 The 12 most-cited domains

The most-cited domain is the marketplace trendyol.com at 28.2% of responses, but what follows is not a marketplace list. Places two, three and four are brand-owned sites belonging to leaderboard entities — karaca.com (17.2%), tac.com.tr (16.9%) and yatasbedding.com.tr (14.6%). Seven of the twelve most-cited domains belong directly to entities in the general leaderboard.

The exceptions are worth naming. eniyilerden.com at #5 (13.7%) and eniyimiz.tr at #12 are Turkish “best of” listicle sites, and kadinlarkulubu.com at #10 (8.5%) is a long-running women’s forum. Editorial and forum sources are present and meaningful in this vertical, but they sit among brand sites rather than above them.

9.2 Source-type mix

9.3 The brand-owned finding

Brand-owned sites account for 52.82% of all citations across 180 domains — more than every other source type combined. The commercial classes together — brand-owned, retailer and marketplace — reach 67.93%. Editorial contributes 10.91% across 23 domains and forum/social 5.32% across 14.

Two things follow. First, in this vertical the assistants are substantially reading what the brands publish about themselves. That is a materially different discovery environment from one mediated by retailer assortments and third-party roundups: a brand’s own site is a source it controls, and here that source is doing most of the work.

Second, the residual is small and genuinely miscellaneous. The 15.61% classed “other” spans 282 domains — editorial and forum sites that fell below the classifier’s keyword lists, plus small independent commerce sites with no tracked entity behind them, such as uysalyorgan.com.tr, homeiswell.com.tr, konforyatak.com.tr and cihanyorgan.com. No small set of intermediaries controls the picture.

Read as a description rather than advice: in this market, visibility appears associated with maintaining a substantial and current own-site presence, more than with placement in third-party assortments. That is not a claim that publishing more will raise a score — it is an observation about where these systems currently read.

10. What the Patterns Suggest

One-sentence takeaway: The visible names share breadth across all five models and a clear product context; the widest gap in the data is not between Turkish and foreign brands, but between the four entities everyone names and everyone else.

This chapter is a neutral reading of the data. It describes patterns associated with visibility; it is not advice, a service, or a recommendation, and visibility remains separate from quality.

1) Breadth is the baseline condition. 19 of the 20 qualified general entities appear in all five models. The single exception, Hepsiburada, sits at #18 despite a 20.44% mention rate, precisely because only four models name it. Broad model coverage is what the visible entities share; an entity strong in one assistant and absent from three does not rank. The pattern repeats at segment level, where the thin qualifiers in Curtains and Towels are almost all breadth-2 entries.

2) Mention and position are different signals. Some entities are named often but late — Hepsiburada (20.44%, MRR 0.151), Madame Coco (26.04%, 0.196), Trendyol (23.38%, 0.225). Others are named rarely but first — Hamam (5.60%, 0.617), Brillant (10.76%, 0.433), Penelope (9.96%, 0.426). Both patterns clear the 5% bar and both are real, but they describe different positions in the market: one is a name that comes up in passing, the other a name that comes up as an answer.

3) Product context determines the leader, not the category. No entity leads more than one of the four segments. Yataş dominates bedding and does not qualify at all in towels or curtains. Chakra leads towels and sits fifth in multi-category. A brand’s visibility here is not a single number; it is a number per product context, and the useful question is which segment a brand owns rather than where it sits overall.

4) The concentration at the top is the structural fact. Four entities — Taç, Özdilek, Karaca, Yataş — occupy the top four places in the general, local, discovery, attribute and use_case cuts, in almost the same order each time, and three of the four appear in the top four of every segment. Below them the leaderboard descends quickly into entities named in a tenth of answers or fewer. Of the 266 tracked entities, only 20 qualify in the general cut and 77 in any cut at all: 189 tracked entities qualify nowhere. The visibility gap in this vertical is not national, it is positional.

5) Domestic visibility is not the problem here — differentiation is. Turkish brands take 89.2% of open-cut mentions without being asked for. But that success is concentrated in a handful of names. For the long tail of Turkish home-textile businesses — and the dictionary holds 232 of them — the relevant gap is not that assistants prefer foreign brands, because they do not. It is that assistants name the same few domestic brands regardless of what is asked. The specialists that do break through do so on narrow, specific questions: İ.D İpek Dokuma on fabric quality, Ecocotton on organic cotton, Hamam on towels, Brillant on curtains.

How a brand can locate itself in the data. Using the dataset (available on request), a brand can read five things in order: whether it appears in all five models; how often it is mentioned within its cut; how early it is named when mentioned; which segment and question type it is strongest and weakest in; and whether it appears at all in the open cut as well as the local one. These readings show where a visibility gap sits, without saying anything about whether its product is good.

11. Limitations and Notes

One-sentence takeaway: This is a single, point-in-time snapshot of one Turkish vertical, scored on a scale that does not travel between editions; read the figures within the limitations below.

• Single collection window. All data comes from a four-hour window on 8 August 2026. Model outputs drift; these figures are a snapshot, not a stable property of the models. This is the first (baseline) edition; quarterly editions will enable trend analysis.
• Scores are not comparable across cuts, editions or verticals. Each component is scaled against that cut’s maximum, so 100 in one cut is a different absolute quantity from 100 in another, and the scale moves whenever the leader moves. Any comparison beyond this edition uses the raw layer — mention rate, MRR and breadth — published for every entity in Chapter 12.
• The score rewards position as well as frequency. This is by design, but it produces counterintuitive placings. Hamam ranks #7 in the general cut on a 5.60% mention rate because its MRR of 0.617 is the highest in the cut. Twelve entities across the ten cuts sit in a top ten on under 10% mentions, including Ecocotton, Clasy, Varol Tekstil, Dali Tekstil, Fullplise, İ.D İpek Dokuma, Penelope and Brillant. §3.2 gives the mention-rate view of the general cut for comparison; read thin-mention ranks cautiously everywhere.
• Curtains is indicative. 7 questions and 175 responses sit below the study’s usual floor, so the 5% threshold is met by 8.75 responses. Nine of its 30 qualifiers are named by fewer than three models, and all three of the study’s sub-10-response qualifiers are here. The top two are robust; the rest of the table should be read as directional.
• Multi-Category has no attribute questions (6 discovery, 6 use_case, 0 attribute). Its leaderboard reflects discovery and use-case phrasing only, and its distinct-entity count must not be compared against other segments.
• Question types are not balanced within segments. Bedding & Quilts carries 8 of the 15 attribute questions; Multi-Category carries none. Discovery prompts reliably surface more distinct entities per answer, so cross-segment comparisons of entity counts partly reflect the design, not the market. Within-segment rankings are unaffected.
• Never compare local against foreign across all 45 questions. The local cut constrains answers to domestic brands by construction. Chapter 8 uses only the 23 open questions, and both denominators are stated wherever an origin figure appears.
• Grounding is not uniform, and one model dominates the ungrounded set. 155 of the 157 ungrounded responses are Gemini’s. Chapter 6 therefore compares Gemini against itself rather than presenting an all-models split, which would measure model identity rather than grounding.
• Out-of-scope entities are excluded, and some of them are large. 30 entities appear in no leaderboard and no origin share: 20 manufacturers and corporate parents, 7 ingredient trademarks and 3 adjacent-market brands. Two of them clear the 5% threshold on volume — Zorlu Holding at 7.56% and Zorlu Tekstil at 6.84% — because models named the parent group instead of a consumer brand. Zorlu is the group behind Taç; naming it in answer to “which sheets should I buy” is not a brand recommendation, which is why these entities are tracked but not ranked. Readers who consider corporate parents part of the visible shelf should treat the leaderboards accordingly.
• Sub-brands and product lines under-count structurally. A product-line alias credits its parent only where the parent is independently identifiable in the same response. 11 orphan observations across 5 entities were recorded and excluded from those entities’ totals rather than credited.
• Origin follows the trademark, which has consequences. Pierre Cardin is classified Foreign despite being operated in Turkey under licence. Talesma is recorded as Turkish on trademark and operator, although its Canadian distribution centre makes a naive registration check disagree; this was verified manually. Two entities carry an unknown origin (Noémie, Thermoclean) and are excluded from every origin share.
• Adjacent entities are included. 63 of the 266 entities are furniture, homeware and fashion brands with genuine home-textile ranges — including Karaca (#3) and IKEA (#9) — retained because the study deliberately asked questions they answer. All tables can be regenerated Core-only from the underlying data.
• Turkish-language prompts only. Results may not transfer to the same questions asked in English about the Turkish market.
• Small score gaps are not meaningful. In the packed middle of each leaderboard, differences of a point or two should not be over-interpreted; only larger gaps are robust. IKEA and Madame Coco are separated by 0.004 points in the general cut.
• Visibility is not quality. This report measures only mention and position. It does not assess whether a brand was described positively, whether a recommendation was accurate, or whether a product is good. A score is not an endorsement.
• Models are probabilistic. The same question can produce different answers; five repeats reduce but do not eliminate this.
• Source-type mix is heuristic. Domains were auto-classified and then attributed to entities by root-website ownership, falling back to a name match on the domain label. Treat the mix as indicative. Source coverage also varies by model.
• Precision over recall. Entity matching was word-boundary anchored and cautious to avoid false positives, so brands mentioned indirectly may be undercounted. There is no sentiment layer; the study is strictly quantitative.
• Output-length limits and model versions differ and date quickly; findings are specific to the versions in Chapter 2 as of August 2026.

Frequently Asked Questions

12. Appendix & Data

12.1 All questions (45) and their categories

Distribution: 15 discovery, 15 attribute, 15 use_case; 22 local and 23 open; across four product segments (18 / 12 / 8 / 7). No question was rewritten for this edition.

12.2 All tracked entities (266)

12.3 Data availability

The study tracked 266 entities; of these, 20 cleared the 5% threshold in the general cut, 22 in the local cut, 22 in the open cut, and 77 in at least one of the ten cuts. To support scrutiny and reproduction, the underlying data is available on request: response-level entity mentions with grounding flags, all ten qualified leaderboards and their brand-only variants, full unfiltered entity metrics, per-model and per-question-type breakdowns, the open-market breakdown, the full 528-domain source list, the entity dictionary with its decisions log, and the methodology and data dictionary.

All figures reflect a single point-in-time run (8 August 2026) and are subject to the limitations in Chapter 11. This is the first (baseline) measurement; the study will be repeated quarterly.

About Herm.io & disclosure

Herm.io is a consumer behaviour and marketing data company. We study how people discover and choose brands to help businesses reach the right customers. This report forms part of our public research and is conducted quarterly.

Disclosure & neutrality

The rankings and metrics in this report are entirely impartial and are based on the methodology described within the report. No brand can pay or provide sponsorship to be included in the report, improve its ranking or influence how it is described in the content.

A brand’s score in the report is not an endorsement or assessment of quality; it is solely a measure of its current visibility within AI models. No service or product offered by Herm.io is sold for the purpose of directly changing the results of this report or guaranteeing a ranking. To preserve objectivity, all references to Herm.io’s own domain are removed from the source data before analysis. This ensures that the company does not appear in, measure itself through or benefit from its own study.

Brands seeking to understand their position in the data can schedule a consultation for an impartial assessment of the findings. The consultation is advisory and free of charge.