--- title: "Pantry to Plate AI Launch Case Study" site: "Silverback Marketing" domain: "silverbackmarketing.com" type: "Case Study" description: "How a brand-new food app scored 82/100 for AI visibility seven days after its domain was registered, with the AI Readiness Kit and 183 Markdown mirrors." canonical: "https://silverbackmarketing.com/case-study/pantry-to-plate" markdown: "https://silverbackmarketing.com/case-study/pantry-to-plate.md" --- # Pantry to Plate launched ready for AI. Silverback built pantrytoplate.food and its apps with the AI Readiness Kit in place from the first deploy. Seven days after the domain was registered, the site scored 82 out of 100 for AI visibility. Fifteen days after that, it scored 99 out of 100 across six areas of site quality. Consumer app · Food & cooking · pantrytoplate.food · Web, Mac, iPhone, iPad, Android · Published November 12, 2026 ## Results at a glance - 82/100 GEO score, Advanced tier, seven days after the domain was registered (October 22, 2026) - 99/100 Six-area site audit score across SEO, content, speed, accessibility, CRO and schema (November 6, 2026) - 0 Critical issues in either audit - 183 Pages in the sitemap, every one with a working Markdown mirror (November 12, 2026) ## Executive summary ### A new product cannot wait years to be understood. Pantry to Plate is a kitchen companion: track what is in your pantry, scan groceries and receipts, save recipes, plan the week and build a shopping list, with optional friend kitchens for sharing. It began with a pantry clean-out, two forgotten bags of pancake mix and a notebook. Silverback built the whole product: the website at pantrytoplate.food and the native apps for Mac, iPhone, iPad and Android. The apps are not in the app stores yet, so at launch the website carried the entire job of explaining the product to people, to search engines and to AI systems. We treated that as an engineering problem. Every page shipped with structured data, a clean Markdown mirror and a place in the AI Readiness Kit file set. Two audits in the first month show what that produced, and they are just as clear about what a new brand still has to earn. ## The challenge ### New domains start with nothing. AI still has to answer. A brand-new site has no links, no reviews and no history. People still ask AI assistants cooking questions every day, and those assistants need something they can read and trust. Questions like these: - "What can I make with what is already in my pantry?" - "How do I use up an opened can of pumpkin?" - "Is there an app that tracks my pantry and plans meals?" - "How do I make a simple caprese?" > Authority takes time and nobody can shortcut it. Readability does not. From the first crawl, every AI system should be able to read the site, see exactly what each page is, and quote it accurately. That part is engineering, and it was ready on day one. ## What we built ### One product. Built to be read by people and machines. The site is prerendered Next.js on Vercel. The audit measured a time to first byte of about 0.20 seconds, and every public page was designed to answer a person first and a parser second. - A website and five apps: The public site plus native apps for Mac, iPhone, iPad, Android phones and Android tablets. Each app has its own support center on the site, so the help content is already published before the apps reach the stores. - Recipes written as data: Every recipe page carries Recipe and HowTo schema: ingredients, numbered steps, prep and cook times, yield and a credited kitchen as author. In the November sample, all 14 recipes checked had every one of those fields. - Nothing asserted that is not true: Rating markup appears only on recipes that have a real rating. Nutrition markup is withheld while ingredient matching is partial, and the page labels its numbers as a partial estimate. The audit called that the correct call. Captured from the live site on November 12, 2026. Public recipes are readable before anyone creates an account, which is what lets search engines and AI systems read them too. ![Pantry to Plate homepage on a desktop browser with the headline Your next great meal starts right here and an example kitchen showing a grocery receipt turned into a stocked pantry](/assets/case-studies/pantry-to-plate/home-desktop.avif) ![Caprese recipe page on a phone, shared by Wittmann Kitchen, with a photo of tomato and mozzarella topped with basil](/assets/case-studies/pantry-to-plate/recipe-mobile.avif) ## The AI Readiness Kit ### The machine layer shipped with the first deploy. Most sites add AI files after launch, if at all. Here they were part of the build. The file set tells crawlers what the product is, which pages exist, what each one answers and how the content may be used: search yes, AI answers yes, model training no. The framework behind it is the [AI Readiness Kit](https://ai.silverbackmarketing.com), Silverback's open toolkit for making a site legible to AI systems. Every item below was confirmed live on November 12, 2026. - llms.txt linking the Markdown mirrors - llms-full.txt extended context - ai.txt usage policy - ai-sitemap.xml alongside sitemap.xml - ai-schema.json and ai-entities.json - RAG index in JSON and JSONL - Read-only OpenAPI description - 28 named AI crawlers addressed in robots.txt - Content-Signal: search yes, AI input yes, AI training no - A Markdown mirror for all 183 sitemap pages - Canonical Link header on every mirror - Recipe, HowTo, FAQPage, VideoObject and SoftwareApplication schema ## Markdown Mirrors ### Every recipe, mirrored for machines. Add .md to any page address and you get the same content as clean Markdown, with a canonical header pointing back to the page people see. On November 12, 2026 we requested the mirror of every URL in the sitemap. All 183 answered with Markdown. For a recipe, that means an assistant gets the title, summary, times, servings, ingredients and steps without parsing a single layout element, and it knows which page to cite. ## Results ### Two audits in the first month. No critical issues in either. ### Launch timeline - October 15, 2026: Domain registered. The WHOIS creation date for pantrytoplate.food. The earliest page date in the sitemap is the next day. - October 22, 2026: GEO audit: 82 out of 100. Advanced tier, up from 80 as fixes landed. 126 URLs in the sitemap, 105 of them recipes, each with a Markdown mirror. No critical issues. - November 6, 2026: Site audit: 99 out of 100. Six areas scored. 42 content pages crawled, every one returned 200 with one H1, a unique title and description, and a self-referencing canonical. Zero P0 or P1 findings, one P2. - November 12, 2026: Live recount. 183 URLs in the sitemap, 150 of them recipes, 10 Journal articles. Every mirror requested returned Markdown. Both audits were run by Silverback, the same team that built the site, using the rubrics we run for any site. Read them as a build that was checked hard and held up, not as an independent award. They measure readiness. They do not measure rankings, traffic or citations, and we are not claiming any here. ### How AI reads the site, dimension by dimension. GEO audit · October 22, 2026. Six weighted dimensions make up the composite. Five sit between 76 and 97. One, brand authority, is where a seven-day-old domain shows its age. Overall: 82 out of 100, Advanced tier. - Technical Foundations: 97 out of 100 (weight 15%) - AI Citability & Visibility: 93 out of 100 (weight 25%) - Structured Data: 91 out of 100 (weight 10%) - Platform Optimization: 78 out of 100 (weight 10%) - Content Quality & E-E-A-T: 76 out of 100 (weight 20%) - Brand Authority Signals: 59 out of 100 (weight 20%) Scores are directional, not a guarantee of AI placement. #### Readiness by AI platform - Google AI Overviews: 82 out of 100 - ChatGPT Search: 80 out of 100 - Perplexity: 74 out of 100 - Google Gemini: 72 out of 100 The recipe corpus is the kind of content these engines surface for cooking and what-can-I-make-with questions. The cap, in the audit's words, is brand newness. ### Six areas of site quality, five perfect scores. Site audit · November 6, 2026. The overall 99 is the rounded mean of the six areas. A 100 here means no deductions in the sampled scope, not perfection. - SEO technical: 100 out of 100 - Content: 100 out of 100 - Speed: 92 out of 100 - Accessibility: 100 out of 100 - CRO: 100 out of 100 - Schema: 100 out of 100 - 42 of 42 content pages returned 200 with one H1, a unique title and description, and a self-referencing canonical - 0 structured data parse failures across all 42 pages - 20 accessibility scans on desktop and mobile with zero rule violations - 14 of 14 sampled recipes carry name, image, author, ingredients, steps, yield and times - FAQ page with 59 question and answer pairs marked up as FAQPage at the October audit - 8 guide videos marked up as VideoObject A scoped audit judgment, not a Google ranking score, forecast or certification. ### A site that kept growing while it was being audited. The same three counts at each checkpoint. The October and November 6 figures come from the audits; the November 12 figures were counted on the live sitemap. - URLs in the sitemap: 126 on October 22, 169 on November 6, 183 on November 12 - Published recipes: 105 on October 22, 138 on November 6, 150 on November 12 - Journal articles: 3 on October 22, 8 on November 6, 10 on November 12 Growth in published pages does not by itself prove indexing or rankings. ### Fast where it counts, with one honest gap. Speed · lab medians, November 6, 2026. Lighthouse lab medians from unique reports, not the fastest run. These are lab observations, not field Core Web Vitals. - Homepage, desktop: performance 94 out of 100, LCP 0.57 s - Homepage, mobile: performance 76.5 out of 100, LCP 5.38 s - Community, desktop: performance 100 out of 100, LCP 0.27 s - Community, mobile: performance 100 out of 100, LCP 0.93 s - Recipe, desktop: performance 100 out of 100, LCP 0.64 s - Recipe, mobile: performance 92 out of 100, LCP 3.31 s Layout shift was 0.003 or lower on every route. ## What is still ahead ### The work a new brand still has to earn. A case study that only shows the wins is an advertisement. These are the two open items, straight from the audits. - Independent authority (59/100): Brand authority is the one dimension holding the composite down. On October 22 the organization linked two owned profiles, with no Wikidata entity, no review or app directory listings and no press. That is typical for a young consumer app and it is the slowest signal to move. The plan: directory and review listings as the apps reach the stores, a wider set of linked profiles, and earned mentions. - Mobile load on two pages (5.38s): The single open finding in the site audit. Median mobile LCP was 5.38 seconds on the homepage and 3.31 seconds on a recipe page, above the 2.5 second good reference. Desktop and the community page are already well inside it. The fix is route-specific profiling of CSS, JavaScript and third-party work. > You cannot buy a reputation on day one. You can make sure that on day one every search engine and every AI system reads you correctly. Pantry to Plate did, and the authority work now builds on a foundation that does not need to be rebuilt. ## Work with Silverback Launching something new? Start it AI-ready. See how AI systems read your site today. Start with an audit and a prioritized plan. [Request an audit](/contact/request-an-audit)