Koskii — AI SEO / AEO (Answer Engine Optimization) Audit

Prepared: 27 August 2026 · Domain: https://www.koskii.com · Method: live technical crawl (raw HTML, not text-extraction which strips <script> tags), robots.txt + machine-readable file checks, and query-level visibility testing via web_search (a directional proxy for Google AI Overviews, which Google states are "rooted in core Search ranking").

Business context source: business-analysis.md (26-section e-commerce analysis completed this session).

---

Executive Summary

Koskii is above average on the technical foundation for AI citation, and below average on content-driven AEO (the part that wins "discovery" and "consideration" queries).

What's working:

• All major AI crawlers (GPTBot, ClaudeBot, Google-Extended, PerplexityBot) are explicitly allowed in robots.txt — no self-blocking.
• A high-quality, well-maintained /llms.txt exists (dated Aug 20, 2026) with a curated collection map, brand description, and sitemap pointer. This is a genuine competitive advantage most Indian D2C ethnic brands lack.
• Product schema is rich: Product + Offer with price, priceCurrency, priceValidUntil, availability, sku, plus BreadcrumbList, MerchantReturnPolicy, OfferShippingDetails. Collection pages carry ItemList, ClothingStore, Organization, GeoCoordinates.
• hreflang is correctly implemented (en-IN, en-US, x-default) across IN and US domains — good for geographic disambiguation.
• A real FAQ page with FAQPage schema (17 Q&A pairs) exists and is extractable.

What's broken or missing:

• Homepage and About page have ZERO <h1>–<h6> heading tags in server-rendered HTML. They are JS-rendered Next.js shells. AI crawlers that don't execute JS (and many do this selectively) see a heading-less page. Collection/product pages are fine (23/17 headings), but the two highest-authority pages are not.
• No aggregateRating / review schema despite the site showing customer testimonials — AI sees every product as "unreviewed." This compounds the known review-system gap from the business analysis.
• No gtin/mpn product identifiers.
• No content hub / blog (every blog-style slug 404s) — so Koskii cannot be cited for the high-intent discovery queries ("best saree brands," "wedding guest outfit ideas," "what to wear to a sangeet") where competitors and generic listicles win.
• No "Where to Buy" page despite being omnichannel (30 stores + Myntra + website) — a clear, citable strength left invisible to AI.
• No /pricing.md, /agents.md, /.well-known/ucp (these are emerging/optional, but pricing.md matters for agentic buying agents).

Visibility gap (query testing):

• Koskii wins its own name (rank #1 on "Koskii sarees," "Koskii store near me," "Koskii reviews").
• Koskii is absent from "best saree brands India," "buy designer lehenga online under 15000," "wedding guest outfit ideas," "online ethnic wear for women" — those are dominated by Biba, Nykaa, Myntra, and generic blogs.
• Reputation risk: negative third-party reviews (MouthShut, IBTimes, Tripadvisor, YouTube "honest review / bad experience") surface prominently for "Koskii reviews." AI Overviews summarizing "Koskii reviews" will likely blend these in. Koskii has no structured, citable positive-review signal to counterweight.

---

1. AI Bot Access (robots.txt) — PASS

Verified live. The following are allowed (each has an explicit Disallow: line that is EMPTY, i.e. not blocked):

User-agent: GPTBot          Disallow:       (allowed)
User-agent: ClaudeBot       Disallow:       (allowed)
User-agent: Google-Extended Disallow:      (allowed)
User-agent: PerplexityBot   Disallow:      (allowed)
User-agent: Perplexity-User Allow: /
User-agent: *               Allow: /  (only /search, /cart, /wishlist, /account$ etc. disallowed)

No AI search bot is blocked. No action needed — this is correct. (Note: CCBot/Common Crawl is not mentioned; leaving it unaddressed is fine since it is training-only and not a citation engine.)

---

2. Machine-Readable Files for AI Agents

File	Status	Verdict
/llms.txt	200 (dated Aug 20 2026, well-structured)	✅ Strong asset — keep current
/robots.txt	200	✅
/agents.md	404	Gap (Shopify auto-generates this; Koskii's headless setup doesn't). Optional but recommended
/pricing.md	404	Gap — matters for agentic buying agents evaluating price programmatically
/pricing.txt	404	Gap (same as above)
/.well-known/ucp	404	Emerging standard — not required yet
/sitemap_agentic_discovery.xml	404	Shopify auto-generates this for agentic discovery; absent here (headless custom build). Noted

Recommendation: Keep llms.txt maintained (it already links all key collections — good). Add a lightweight /pricing.md with tier/range pricing (the business analysis gives the ranges: sarees ₹1,745–₹39,990, median ~₹5,400; lehengas ₹5,000–₹39,990, etc.). This directly helps autonomous buying agents and is trivial to maintain.

---

3. Schema Markup — Verified via RAW HTML (not text-extraction)

Per the skill's explicit warning, schema was checked with curl + grep against raw <script type="application/ld+json"> so <script> tags were not stripped.

3a. Product page (/products/...-gcss0017471_...)

ld+json blocks: 6. Types present:

• Product ×2 (with sku ✓, Offer ✓)
• Offer ×2 — price ✓ ("12490.00"), priceCurrency ✓ (INR), priceValidUntil ✓, availability ✓
• BreadcrumbList ✓, Brand ✓, MerchantReturnPolicy ✓, OfferShippingDetails ✓

Missing on product schema: aggregateRating (0), review (0), gtin (0), mpn (0). → AI engines read every Koskii product as unreviewed and unidentified by GTIN. Critical gap.

3b. Collection page (/collections/lehengas, /collections/sarees)

ld+json: 10. Types: CollectionPage, ItemList (26 ListItem), ClothingStore, Organization, PostalAddress, GeoCoordinates, ContactPoint, BreadcrumbList. → Strong. H1 present ("Sarees"), 23 headings, server-rendered product names + meta description. Good AI-extractable collection pages.

3c. Homepage (/)

ld+json: 6 (WebSite, WebPage, Organization, SearchAction, EntryPoint). But 0 <h1>–<h6> tags in static HTML (JS-rendered). Title tag present.

3d. About page (/pages/about-us)

ld+json: Org/WebPage only. 0 <h1>–<h6> tags in static HTML. Body <p> text IS server-rendered (103 <p> tags), so prose is readable, but there is no semantic heading hierarchy for crawlers to anchor topic structure.

Schema scorecard

Schema	Product	Collection	Home	About	FAQ
Product / Offer	✅	—	—	—	—
aggregateRating	❌	—	—	—	—
review	❌	—	—	—	—
BreadcrumbList	✅	✅	—	—	—
FAQPage	—	✅*	—	—	✅
Organization	—	✅	✅	✅	—
H1 heading (static)	✅	✅	❌	❌	❌

\* Collection /collections/sarees carries FAQPage schema (2 blocks).

---

4. Content Extractability Audit

Check	Home	Product	Collection	About	FAQ	Customer-Reviews
Clear definition in first para	—	partial	✅	✅ (prose)	✅	partial
Self-contained answer blocks (40–60w)	❌ (JS)	thin	✅	✅	✅	narrative only
Statistics with sources	❌	❌	❌	brand-stated	❌	❌
Comparison tables (X vs Y)	❌	❌	❌	❌	❌	❌
FAQ section	❌	❌	partial	❌	✅	❌
Expert attribution	❌	❌	❌	founders named	❌	testimonials, unnamed
Last-updated signal	n/a	❌	❌	❌	❌	❌
Heading hierarchy (static)	❌	✅	✅	❌	❌	✅ (prose)

Extractability verdict: Product and collection pages are reasonably extractable. Homepage, About, and FAQ pages are JS/heading-light and miss freshness + stat/attribution signals. There are no comparison tables and no statistics blocks anywhere — both are top GEO visibility boosters (+37–40% per Princeton KDD 2024).

---

5. AI Visibility Query Testing

Tested via web_search (top-5). Directionally mirrors what Google AI Overviews and AI assistants cite, because Google's generative features are built on core ranking.

Query	Koskii cited?	Who ranks instead
Koskii sarees official store	✅ #1	(Koskii)
Koskii store near me Bangalore	✅ #1 (store-locator)	Koskii, magicpin, Tripadvisor
Koskii reviews customer feedback	✅ #1 (reviews) + ⚠️ negatives #2–5 (MouthShut, IBTimes, Tripadvisor, YouTube)	mixed — reputation risk
best saree brands India online	❌	UT Sarees, Rashika Mittal, Rana's, Mavuris, Myntra
buy designer lehenga online India under 15000	❌	Aachho, Myntra, Flipkart, Indiamart
Koskii vs Kalki bridal wear	❌ (no direct comparison asset)	Kalki-film noise + Tripadvisor
wedding guest outfit ideas women ethnic wear	❌	NDTV Shopping, Beauty&Flowers, VilleFashion
online ethnic wear for women India	❌	Andaaz, House of Indya, Biba, Nykaa, Anandi

Interpretation:

• Brand/navigational: strong (owned-name queries = #1).
• Discovery / consideration / comparison: weak-to-absent. These are exactly the queries where AI answers consolidate a short list of cited brands. Koskii is invisible there, so when a user asks ChatGPT/Perplexity "what are the best saree brands in India?" Koskii is not in the answer.
• Reputation: negative third-party sentiment already outranks Koskii's own positive-review page for "Koskii reviews." AI Overviews will likely surface this mix. Koskii needs a structured, citable positive-review signal (see recommendations).

---

6. Priority Recommendations

Ordered by impact × effort.

P0 — Fix before anything else

1. Add aggregateRating + review JSON-LD to every product page. Even a modest "Rated 4.x from N customers" block (backed by real review data) flips 14,000+ products from "unreviewed" to "reviewed" in structured data. This is the single highest-leverage AEO fix and also a known CRO/trust gap. If reviews aren't collected yet, prioritize the review-collection system first (business analysis gap #3).
2. Restore <h1> (and a basic heading hierarchy) to the homepage and About page in server-rendered HTML. Today those pages have 0 static heading tags — weak for non-JS AI crawlers and for accessibility. Collection/product pages prove the theme can server-render headings; apply the same to home/about.

P1 — High impact, build content

1. Launch a content hub (blog/style-guide) targeting discovery queries. The biggest visibility gap is absence from "best saree brands," "wedding guest outfit ideas," "what to wear to a sangeet/haldi." Publish definitive, extractable guides with:
   • Definition blocks (40–60 words) answering "what is [occasion] wear?"
   • Comparison tables (e.g., "Lehenga vs Saree vs Gown for a wedding guest")
   • Statistics with sources (e.g., "Indian occasion-wear market ~₹35,000 Cr — YourStory 2026")
   • FAQ sections with FAQPage schema (reuse the existing FAQ pattern)
   • Author attribution (founder/stylist name + title)
   • "Last updated" dates
2. Create a "Where to Buy" page listing all channels: koskii.com, us.koskii.com, 30 stores (8 cities), Myntra, app. Link each. Captures high-intent "where to buy Koskii" and reinforces the genuine omnichannel strength (currently invisible to AI).
3. Build comparison/versus content ("Koskii vs Kalki," "Koskii vs Biba," "Designer lehenga under ₹15k") with balanced comparison tables. Currently "Koskii vs Kalki" returns only noise; owning that query with citable, structured comparison is low-cost and high-value.

P2 — Authority & trust signals

1. Add gtin/mpn to product schema where available (catalog has proprietary SKUs like GCSS0017471; map these to mpn).
2. Counter the reputation risk. Proactively seed structured, citable positive signals: verified-buyer review snippets with star counts in schema, and a "As featured in" / press section (Inc42, YourStory, Tribune India) with links. When AI summarizes "Koskii reviews," real structured positives should dilute the third-party negatives.
3. Add a lightweight /pricing.md (tier/range pricing) for agentic buying agents. Include "Last updated" date.

P3 — Optional / emerging

1. Add /agents.md (agent instructions) — Shopify auto-generates this for standard stores; Koskii's headless build doesn't have it. Optional but cheap.
2. Monitor AI visibility monthly: re-run the 8 queries above across ChatGPT/Perplexity/Google and log citation rate. Tools: Peec AI, Otterly, ZipTie (cross-platform) — or manual monthly check.

---

7. What NOT to Do (per Google's AI optimization guide)

• Do not write separate "for AI" content or chunk pages into AI-bait fragments — risk of scaled-content-abuse spam action. The content hub above serves people AND AI with normal structure.
• Do not block GPTBot/ClaudeBot/Google-Extended (they're already allowed — keep them).
• Do not fabricate review counts/stars. Only surface real review data in aggregateRating.
• Do not gate the content hub behind login — AI can't cite what it can't read.

---

8. Sources & Verification Notes

• robots.txt, llms.txt, schema, and heading counts: directly verified via curl raw HTML on 2026-08-27 (homepage, /collections/sarees, /collections/lehengas, a live product URL, /pages/about-us, /account/faq, /account/store-locator).
• Schema absence claims (aggregateRating/review/gtin) verified by grepping raw <script type="application/ld+json"> content — NOT inferred from text-extraction (which strips scripts).
• Query visibility: tested via web_search top-5 on 2026-08-27; Google AI Overviews correlate with organic ranking, so this is a directional proxy. ChatGPT/Perplexity direct citation testing could not be run in this environment — recommend manual re-test on those platforms.
• Business context: business-analysis.md (this session) — 14,227 SKUs, 602 collections, 30 stores/8 cities, ₹61 Cr Series A, median price ~₹5,400, no structured review system, English-only.

Confidence: [High] for all technical findings (directly observed). [Medium] for query-visibility interpretation (proxy method). [Medium] for competitor-citation claims (based on observed SERP, not live AI answers).