A Real AI and SEO Audit, Page by Page | Content Cucumber

A real Content Cucumber AI and SEO audit, in full. The exact format, depth, and live four engine AI citation scoring a client receives, run on a fast commerce site that scores zero in the AI answer layer. Read it here or download the PDF.

Source: https://contentcucumber.com/research/ai-seo-audit-sample/

By Brent Peterson, CEO, Content Cucumber. Updated 2026

Sample audit

A complete Content Cucumber audit of a fast, well built commerce site that scores zero in the AI answer layer. The exact format, depth, and four engine AI citation scoring a client receives. The whole audit is on this page, and the PDF needs no form.

Download the PDF

Cucumber & Co.

A Real AI and SEO Audit, Page by Page

SEO + AI AUDIT · INTERNAL DEMO SHOWCASE

A comprehensive audit of the Content Cucumber commerce demo (shopware.contentcucumber.com), a fictional 1826 B2B cucurbit and brining-stock trade catalog. It is fast, well-built, and beautifully written, and it is completely invisible to the AI answer layer. That contrast is the entire point of the demo. AUDIT DATE 2026-05-17 shopware.contentcucumber.com homepage as captured 2026-05-17. Audit findings reference this snapshot.

The short version

https://shopware.contentcucumber.com/ AUDITED URL 2026-05-17 AUDIT DATE Internal (Content Cucumber commerce demo: "Cucumber & Co.", fictional B2B trade catalog) PREPARED FOR Content Cucumber PREPARED BY Comprehensive. External, public-data only. PSI mobile + desktop. Live AI citation baseline across AUDIT TYPE OpenAI, Gemini, Perplexity, and Claude (20 queries, two tiers). Internal demo showcase. No pricing. This is our own demo; the closing is a remediation plan, not an FRAME offer. This is a genuinely good storefront that no AI engine can see. That is the lesson, and it is the cleanest

possible illustration of the Content Cucumber thesis on a commerce site.

Cucumber & Co. is a serverrendered Astro build behind Caddy, fast (desktop Performance 98, LCP 0.9s; mobile 86, CLS 0), with a real product catalog (five categories, twelve product pages), a clean information architecture (heritage, growers, sustainability, press, EDI, allocations, quote), a strong human-written title and meta description, an open robots.txt with sensible commerce disallows, and a complete sitemap. The craft is real. The copy ("a long, slow grudge against bitterness") is on-brand and excellent. And the machine-readability layer is essentially absent.

Zero structured data anywhere

(no Organization, no Product, no Offer, no BreadcrumbList on a commerce site), no Open Graph or Twitter cards at all, and llms.txt is a 404. The live four-engine test confirms the consequence: across twenty wholesale-cucurbit category queries, Cucumber & Co. holds

0% share of voice on OpenAI, Gemini, Perplexity, and Claude

, with zero brand mentions. Composites land at OpenAI 21, Gemini 52, Perplexity 41, Claude 21, carried entirely by generic sentiment when the name is supplied. A buyer asking an engine for a wholesale cucumber supplier is handed Sysco, US Foods, and Gordon Food Service. Cucumber & Co. does not appear. Some of that is by design (a fictional 1826 brand has no real-world authority footprint, and never will). But the technical gaps are real and would apply identically to a live commerce client: a fast, attractive store with no schema, no Open Graph, and no llms.txt is a store the AI answer layer cannot describe, cannot enrich, and will not cite. The demo is the argument. BRENT'S TAKE This is the best teaching tool in the set because it is ours and we built it well on purpose. The store is fast, the copy is sharp, the catalog is real, and the AI layer still returns nothing, zero of twenty queries, zero share of voice on all four engines. Part of that is a fictional brand with no authority, which is honest to say. But the technical half is the part that maps straight onto a real merchant. No Product schema on a product site. No Open Graph, so every shared product link is a bare URL. No llms.txt. A beautiful store the engines cannot read is still invisible. That is the whole pitch, and now we can show it instead of explain it.

SEO and AI Health Score

Categorical scores weighted by impact. PSI mobile + desktop captured 2026-05-17 via Google PageSpeed Insights API. AI Search scored on measured live citation outcome. OF 100

Category breakdown

Crawlability and Indexing URL Structure and On-Page Schema and Structured Data Performance and Core Web Vitals Social and Open Graph Mobile Content Depth AI Search and GEO (infrastructure) AI Search and GEO (measured outcome)

Lighthouse PSI (2026-05-17)

Page / Strategy Performance mobile

/ desktop

/ Performance is strong: zero TBT both strategies, CLS effectively zero, desktop near-perfect. Mobile LCP 3.1s is the one soft metric (the "good" line is 2.5s), driven by FCP 2.9s; a hero-image preload closes it. This is a fast, well-engineered front end. The score is held down by machine-readability and AI visibility, not speed.

80% 64% 8% 84% 12% 80% 60% 38% 12%

SEO (Lighthouse) LCP FCP CLS TBT

100 3.1s

2.9s 0 0ms

100

0.9s 0.9s 0.005 0ms

What this comprehensive audit covers

The full nine-part deliverable, run here on our own commerce demo. 1. Score and category Out-of-100 health score weighted by impact, AI category split into infrastructure and breakdown measured outcome. 2. Tech stack What the site reports: platform, edge, analytics, schema, sitemap, llms.txt, content surface. 3. Top findings The gaps that hold the score down and what each would add if shipped. 4. Per-check findings detail Pass / Partial / Fail / Manual per check across the canonical categories. 5. Upgrade code Copy-paste blocks for the machine-readability gaps (Product / Organization schema, Open Graph, llms.txt). 6. Live AI citation tests 20 category queries, two tiers, run live on OpenAI, Gemini, Perplexity, and Claude on 2026-05-17. 7. 90-day roadmap What we would sequence on this demo to make it a complete reference build. 8. Internal remediation plan This is our own demo. No pricing. The fixes we would ship and why. 9. What Content Cucumber The production model that builds the authority surface the demo lacks. produces

Tech stack

What shopware.contentcucumber.com reports to a direct request (raw HTML, headers, robots, sitemap, llms.txt, route enumeration) on 2026-05-17. Platform Astro (static site generator; KB home. Edge / server Caddy ( Analytics / tools None detected in static HTML. No GTM, GA4, or pixel. Structured data

None.

BreadcrumbList. On a commerce site this is the single biggest miss. Open Graph / Twitter

Absent.

renders as a bare URL. Title / meta description Strong. Title "Cucumber & Co. · Cucurbits and Brining Stock Since 1826"; a real, human-written meta description. robots.txt Present and well-formed. ( /cart directive. sitemap.xml Present and complete. ~28 URLs: home, /products with five category facets, twelve product detail pages, and the heritage/growers/sustainability/press/EDI/allocations/quote/contact set. llms.txt

404.

Content surface A real catalog (five categories, twelve products) plus a genuine brand-story surface (heritage, growers, sustainability, press). No blog or editorial authority layer. THE INSTRUCTIVE PART This is a fast, server-rendered, well-structured commerce site with a real catalog and good copy, and it carries no schema, no Open Graph, and no llms.txt. It is the exact profile of a real merchant who invested in design and speed and skipped the machine-readability layer. Section 6 shows what that costs in the AI answer layer: nothing the engines can see, so nothing they cite. assets). Server-rendered HTML, ~17 /_astro/ ). HTTPS. via: 1.1 Caddy Zero JSON-LD anywhere: no Organization, no Product, no Offer, no No or tags. Every shared product or category link og: twitter: with sensible commerce disallows Allow: / , , , , ) and a Sitemap /checkout /account /admin /api No curated entity/catalog map for language models.

Top findings

1. ZERO STRUCTURED DATA ON A COMMERCE SITE No JSON-LD anywhere. For a store with a twelve-product catalog and five categories, the absences that matter are Organization, Product + Offer (price, availability, SKU), and BreadcrumbList. Without Product schema there is no rich-result eligibility, no price/availability surface for shopping experiences, and no machine-readable entity for AI engines to ingest. This is the single highest-leverage fix on the site.

What it would add.

Rich-result eligibility, a clean entity for AI ingestion, and a price/availability layer commerce surfaces read. Copy-paste Organization + Product blocks in section 5. 2. NO OPEN GRAPH OR TWITTER CARDS ANYWHERE No or tags. For a B2B catalog whose buyers share product and category links in email, og: twitter: Slack, and LinkedIn, every shared link renders as a bare URL with no title card, image, or description. This depresses click-through on exactly the referral paths a trade catalog runs on.

What it would add.

Proper unfurls on every shared product and category link. One head block, sitewide. Section 5. 3. NO LLMS.TXT, AND 0% AI SHARE OF VOICE llms.txt is a 404, and the live four-engine test returns 0% share of voice with zero brand mentions on all of OpenAI, Gemini, Perplexity, and Claude. Part of this is structural (a fictional 1826 brand has no real authority footprint), but the technical half is real: with no schema and no llms.txt, even a crawlable store gives the engines nothing curated to map. On a live merchant this is the difference between being ingestible and being invisible.

What it would add.

A curated catalog map for the crawlers, plus the schema in finding 1, makes the store fully ingestible. Authority (the part a real brand earns) is then the content job. Section 6 and 7.

On-Page and Structured Data

Crawlability, Indexing, and AI Access

If Check Score broken 1.1 robots.txt PASS accessible 1.2 sitemap.xml PASS present 1.3 llms.txt Medium FAIL 1.4 AI crawler access PASS 1.5 Server-rendered PASS content 1.6 HTTPS PASS

If Check Score broken 2.1 Title tag PASS 2.2 Meta description PASS 2.3 H1 PASS 2.4 Open Graph / High FAIL Twitter 2.5 Organization High FAIL schema 2.6 Product / Offer Critical FAIL schema Finding Present, well-formed. with sensible commerce disallows and Allow: / a Sitemap directive. Present and complete. ~28 URLs including all twelve product pages and category facets. 404. No curated catalog/entity map for LLMs. Section 5. /llms.txt Open robots, no AI-bot blocks. Crawlers can reach everything; there is just little machine-readable structure to ingest. Astro static HTML at first contact. No JS-render blind spot. HTTPS via Caddy. Correct. Finding "Cucumber & Co. · Cucurbits and Brining Stock Since 1826". Distinctive, category-anchored. Real, human-written, on-message. Done right. Single H1, strong voice ("a long, slow grudge against bitterness"). On-brand. None at all. Shared product/category links render bare. See finding 2; block in section 5. None. No machine-readable entity for the business. Section 5. None on any of the twelve product pages. No price, availability, or SKU surface. The highest-leverage gap on a commerce site. If Check Score broken 2.7 BreadcrumbList / Low MANUAL canonical

Performance and Core Web Vitals

If Check Score broken 3.1 Desktop vitals PASS 3.2 Mobile PASS performance 3.3 Mobile LCP Low PARTIAL 3.4 CLS PASS 3.5 TBT PASS 3.6 SEO PASS (Lighthouse) Finding Category facets exist ( ); confirm self-referential /products?cat= canonicals and add BreadcrumbList with the schema pass. Finding Performance 98, LCP 0.9s, FCP 0.9s. Near-perfect. 86 of 100. Strong for a content-rich storefront. 3.1s, just over the 2.5s "good" line, driven by FCP 2.9s. Hero-image preload closes it. 0 mobile, 0.005 desktop. Excellent layout stability. 0ms both strategies. Minimal JS, as expected from a static Astro build. 100 both. On-page basics are correct (the shallow check does not see the schema/llms.txt gaps).

D. /llms.txt (real text file at root)

Upgrade code

Copy-paste for the three machine-readability gaps. These are the corrective fixes, not marginal polish.

A. Organization JSON-LD (site head)

{ "@context": "https://schema.org", "@type": "Organization", "name": "Cucumber & Co.", "url": "https://shopware.contentcucumber.com/", "foundingDate": "1826", "description": "Wholesale cucurbits and pickle brining stock for pickle plants, foodservice distributors }

B. Product + Offer JSON-LD (product page template)

{ "@context": "https://schema.org", "@type": "Product", "name": "{{product_name}}", "category": "{{category}}", "brand": {"@type":"Brand","name":"Cucumber & Co."}, "offers": { "@type": "Offer", "priceCurrency": "USD", "price": "{{price}}", "availability": "https://schema.org/InStock", "url": "{{product_url}}" } }

C. Open Graph + Twitter (site head)

Cucumber & Co. > Wholesale cucurbits and pickle brining stock since 1826. > Trade catalog for pickle plants, foodservice distributors, > and retail buyers. ## Catalog - Products - Pickling - Brining Stock ## Company - Heritage - Growers - Contact A through D are an afternoon on an Astro build. They make the store fully ingestible and eligible. They do not, by themselves,

manufacture the real-world authority a live brand earns through content; that is section 7.

AI search visibility (live, 4 engines)

20 wholesale-cucurbit category queries (10 strategic, 10 marketplace) run live on 2026-05-17 across OpenAI, Gemini, Perplexity, and Claude. This is what a brand with no authority footprint looks like to the AI answer layer.

Per-engine answer-engine score

OpenAI Gemini gpt-5-search · live web 2.5 Pro · live web

21 52

/100 /100 Not in the conversation. Zero Generic description only. Zero share of voice. share of voice. Brand Recognition 1/20 Brand Recognition Market Score 0/10 Market Score Presence Quality 0/20 Presence Quality Brand Sentiment 20/40 Brand Sentiment Share of Voice 0/10 Share of Voice THE PATTERN Four engines,

0% share of voice in every tier, zero brand mentions

carried by generic sentiment when the name is in the prompt; brand recognition is 1 to 8 of 20. Asked for a wholesale cucumber or brining-stock supplier, the engines return Sysco, US Foods, Gordon Food Service, and Performance Food Group. Cucumber & Co. never appears. Some of that is the fictional brand having no authority footprint, which is honest. The transferable half is technical: a fast, attractive store with no schema and no llms.txt gives the engines nothing structured to read, so even ingestion adds nothing citable.

What the engines return instead, strategic tier

Cucumber & Co. holds

0% of every pie below

cucurbit and brining supply. OpenAI Gemini

0% 0%

Cucumber & Co. share Cucumber & Co. share Sysco 60% Sysco 50% US Foods 20% Performance Food Group 30% Gordon Food Service 20% US Foods 20% Perplexity Claude sonar · live web training data · not live web

41 21

/100 /100 Not surfacing. Zero share of No training-data footprint. voice. Zero share of voice. 8/20 Brand Recognition 4/20 Brand Recognition 1/20 3/10 Market Score 2/10 Market Score 0/10 16/20 Presence Quality 15/20 Presence Quality 0/20 25/40 Brand Sentiment 20/40 Brand Sentiment 20/40 0/10 Share of Voice 0/10 Share of Voice 0/10 . The composites (21, 52, 41, 21) are . The slices are the incumbents the engines name for wholesale Perplexity Claude

0% 0%

Cucumber & Co. share Cucumber & Co. share Sysco 55% Sysco 45% US Foods 25% US Foods 30% Gordon Food Service 20% Gordon Food Service 25%

Narrative themes the engines associate with the category

What the engines talk about for wholesale cucurbit and brining supply. Cucumber & Co. is absent from all of it; these are the themes a real brand's content would have to own. OpenAI Gemini National distribution scale and Bulk and commercial-grade reliability supply Foodservice and broadline Specialty and organic sourcing supply Supply-chain consistency Consistent year-round volume Foodservice distributor Established incumbents (Sysco, relationships US Foods) Volume pricing Cost and contract terms Perplexity Claude Premium and artisan sourcing No training-data knowledge of Cucumber & Co. Custom brining blends Defaults to broadline Sustainability and grower distributors relationships Names Sysco, US Foods, GFS National vs regional supply instead Vetted, established suppliers Cannot describe the brand or catalog Zero footprint in the model's memory

90-day roadmap (to make this a complete reference build)

Weeks Workstream Outcome 1 Machine-readability Organization + per-product Product/Offer schema, BreadcrumbList, full Open Graph + layer Twitter block, real llms.txt (section 5 A-D). The store becomes fully ingestible and richresult eligible. 1 Performance polish Hero-image preload to bring mobile LCP under 2.5s. Everything else is already strong. 2-12 Authority surface (the Add the editorial layer a real merchant would need: grower stories, brining technique demo's missing half) guides, sourcing and sustainability explainers, all named-author and dated. This is what would move share of voice off zero if the brand were real. 13 Re-baseline Re-run the 20-query four-engine test + PSI. Track the schema/OG/llms.txt deltas with the new cc-audit-delta.sh

Internal remediation plan

This is our own demo, so there is no pricing and no offer. This is the fix list and the reason the demo is built the way it is. capability against this 2026-05-17 baseline.

Ship now (corrective)

Fix Why Product + Organization + The biggest gap on a commerce site. Rich-result eligibility + a Breadcrumb schema clean entity for AI ingestion. Open Graph + Twitter block Every shared product link currently renders bare. One head block, sitewide. Real /llms.txt Curated catalog map for the crawlers. Currently a 404. Mobile hero preload Closes the one soft performance metric (LCP 3.1s to under 2.5s).

Why the demo is built this way

Choice Rationale Fast, beautiful, well-written front So the contrast is unmistakable. The store is genuinely good and still scores 49 and end 0% AI visibility. Design and speed are not the gap. Fictional 1826 brand Honest demonstration that authority is earned in the real world over time. No schema trick manufactures it. Schema/OG/llms.txt deliberately So a before/after re-audit with absent at baseline machine-readability layer is worth, the same way the aragrow re-audit did. THE POINT OF THE DEMO A fast, attractive, well-built commerce site can still be invisible to the AI answer layer. The technical half of that (schema, Open Graph, llms.txt) is a one-afternoon fix and the demo proves the cost of skipping it. The other half, real authority, is the content work, and it cannot be faked, which is exactly why a fictional brand scores zero no matter how good the store looks. Effort Afternoon (Astro template) Under an hour Under an hour Under an hour shows exactly what the cc-audit-delta.sh

A beautiful store. Invisible to the AI.

Cucumber & Co. is fast, well-designed, and well-written, and it holds 0% share of voice across OpenAI, Gemini, Perplexity, and Claude with zero brand mentions. The technical half of that gap, no schema, no Open Graph, no llms.txt, is an afternoon of work and the demo exists to prove what skipping it costs. The other half, real authority, is earned through content over time and cannot be faked, which is why a fictional 1826 brand scores zero no matter how good the storefront looks. The demo, in three lines. 1.

Speed and design are table stakes, not visibility.

Machine-readability is a one-afternoon fix.

purpose. 3.

Authority is the content, and it is the only half that cannot be shortcut.

Cucumber produces. This is the reference build. Re-baseline with cc-audit-delta after the fixes ship.

What Content Cucumber produces

The proof. The named-author, schema-wrapped, citable content cadence that builds the authority a storefront alone cannot.

Case studies

Named-client case studies built from interviews, structured, schema-wrapped, written to be cited. The asset class that compounds into category authority. contentcucumber.com/our-work contentcucumber.com/services/we-write-casestudies The demo proves the thesis on commerce: a great store the AI cannot see is still invisible. Schema is the afternoon. Authority is the work. This store has both and still scores 49. Schema, Open Graph, llms.txt. The demo skips it on That is what Content

Long-form and research

Industry analysis and thought-leadership in a namedauthor voice. The depth-and-frequency mechanism that moves a brand off zero in the AI answer layer. contentcucumber.com/blog

Questions about this document

What am I looking at?

A complete Content Cucumber AI and SEO audit, run on our own commerce demo site, Cucumber and Co. It is the exact format, depth, and live four engine AI citation scoring a client receives.

Why audit your own demo site?

Because it is fast, well built, and beautifully written, and it still scores zero in the AI answer layer. That contrast shows what a storefront alone cannot do, and what the missing pieces cost.

Can I download the PDF without a form?

Yes. The download button links straight to the PDF version of this audit.

What does an audit like this cover?

Crawlability and indexing, on-page and structured data, performance and Core Web Vitals, social and Open Graph, mobile, content depth, AI search infrastructure, and a live citation test across OpenAI, Gemini, Perplexity, and Claude. Each check comes with a finding and a fix.

Can I get one for my site?

Yes. The sample audit page lets you request one with your URL, and a real person on our team runs and reviews every audit before it goes out.