1. Background and context
For years the industry narrative oscillated between two extremes: "SEO is alive and well" and "SEO is dead because AI search wipes out organic traffic." I used to think the answer was straightforward — make great content and you'll win. Then we tested an experiment with a mid-sized recipe blog (we'll call it SimplePantry) and one single query: "5-step lasagna recipe for beginners." That moment changed everything about the "is SEO dead" debate — not because SEO vanished, but because the rules of visibility evolved.
SimplePantry had been around five years, publishing 400+ recipes. Monthly organic traffic was stable at ~180k users. The site had good on-page basics and a loyal audience. But when Google announced and rolled out Search Generative Experience (SGE) features, visibility became less about traditional snippet chasing and more about how to be the concise, authoritative source that an AI could synthesize into a snapshot.
2. The challenge faced
Objective: Earn the SGE snapshot (the AI-generated answer card) for "5-step lasagna recipe for beginners" and drive meaningful engagement back to the site.
Constraints and friction points:
- Keyword intent is highly transactional and instructional; users want a quick, reliable method. SGE favors concise, authoritative answers with verifiable signals — new territory for many recipe blogs relying on long-form content. Potential for "zero-click" — the snapshot could answer everything users need without clicking through, hurting downstream ad revenue and engagement metrics. Technical: the site used a legacy theme with moderate Core Web Vitals issues and no explicit Recipe JSON-LD for step markup.
Risk: Optimize for the snapshot and potentially lose long-form readers, or ignore it and watch competitors get the AI 'trust' signal that drives impressions.
3. Approach taken
We designed a two-track strategy: optimize a focused, single-purpose page to win the snapshot while keeping a richer long-form version to capture engagement and retain brand depth.
Key principles guiding the approach:
- Answer-first design: present the 5 steps prominently and plainly — think "flash card" for the AI. Structured signals: explicit Recipe schema and step markup so the machine reading layer can extract discrete steps and ingredients. Provenance and credibility: show experience indicators (author notes, testing count, prep photos, and precise timings) to meet E-E-A-T signals. Progressive disclosure: let the snapshot-friendly content live above the fold, and tuck supporting detail, FAQs, tips, and variants beneath for readers who want depth. Instrumentation: define measurement — impressions, snapshot inclusion rate, clicks, session duration, and conversion (newsletter sign-ups / print recipe downloads).
Why this approach?
Analogies help: think of the SGE snapshot as an appetizer served by the search engine. If you provide a neat, high-quality amuse-bouche (clear 5-step recipe), Google will often use it. But you still want the diner to order the main course — the longer article — so you must make the appetizer delicious and link it clearly to an irresistible entrée.
4. Implementation process
Timeline: 12 weeks from hypothesis to measurable outcomes.
Week 1–2: Audit and hypothesis
We ran a content and technical audit. Findings included:
- No structured "step" markup in Recipe schema — only basic recipeCard snippets existed. Page layout buried key steps below a long intro and multiple large images. Mobile LCP averaged 3.2s and CLS spiked because of late image loading.
Week 3–5: Rebuild the target page (A variant) with snapshot-first UX
We created a clean "5-step lasagna" page variant that followed these specifics:
- Top-of-page "5-step at-a-glance" card: Each step numbered, 1–2 short sentences, and one line for timing per step. Prominent ingredient list formatted as bullets, minimum interruption, and exact quantities (no vague "to taste"). JSON-LD Recipe with explicit step objects, cookTime, totalTime, nutrition, and image array. We also included authorExperience metadata (e.g., "Tested 10 times, yields 6 servings"). Schema FAQ block with 6 targeted questions (e.g., "Can I freeze this lasagna?", "Vegetarian swap options?"). Server-side rendering for the critical card to guarantee fast First Contentful Paint (FCP).
Week 6–8: Optimize trust signals and visuals
We added these credibility boosters:
- Step photos shot with consistent lighting and labeled images (image alt text used natural language signals). Author bio that included culinary credentials and "experience" notes (number of times recipe tested, source provenance). User ratings placed near the top with micro-interactions for quick feedback.
Week 9–10: Technical improvements and CRO wiring
Performance upgrades were implemented:
- Lazy-loading non-critical images, preloading hero image, and compressing assets to drop LCP to 1.8s. Reduced CLS by reserving image dimensions and removing layout-shifting third-party widgets. Added clear micro-CTAs: "Print recipe", "Save to cookbook", "Subscribe for weekly recipes" — instrumented in GA4 and server logs.
Week 11–12: Monitoring and iteration
We monitored via Google Search Console (GSC), third-party rank trackers, GA4 events, and server logs. Iterations were small: tweak step wording, adjust FAQ phrasing to match natural-language queries, and refine schema to include 'recipeCategory' and 'recipeCuisine' tags.
5. Results and metrics
Outcome: Within six weeks of the snapshot-first page going live, SimplePantry captured an SGE snapshot for "5-step lasagna recipe for beginners" in approximately 62% of queries for that exact phrase in our tracked regions.
Metric Baseline (prior 8 weeks) After snapshot-first implementation (weeks 6–12) Impressions for target query 4,200 28,600 (+581%) SGE Snapshot inclusion rate 0% 62% Click-through rate (for impressions with snapshot) 5.2% 18.4% (+254%) Organic sessions to page 1,080 4,320 (+300%) Average session duration 1:12 2:38 (+124%) Newsletter sign-ups attributed to page 12 64 (+433%)Interpretation: The snapshot helped drive impressions and, crucially, a higher-quality click rate. The at-a-glance card satisfied many users immediately, but it also acted as the trailer convincing hungry users to visit for the yeschat.ai full recipe, tips, and printable formats. The myth that AI answers always reduce clicks proved simplistic — the right design turned the snapshot into a conversion funnel rather than a rival.
6. Lessons learned
Based on the experiment, here are the distilled lessons — presented like kitchen rules.
- Serve the answer first, then the story: If you hide the core steps behind a long intro, both humans and AI will ignore you. Make the 5-step answer obvious and machine-readable. Structured data is the mise en place of modern SEO: Recipe JSON-LD and explicit step markup are not optional. They are the ingredients that let the AI extract your content cleanly. Experience matters: The new E-E-A-T landscape rewards demonstrable experience. Saying "tested 10 times" and showing consistent photos is stronger than vague authority claims. Performance is visibility: Fast critical rendering and stable layout increase the chance the AI uses your content for a snapshot — think of it as being able to deliver a clean plate quickly in a busy kitchen. Design for both the machine and the diner: The snapshot-first card must be concise for AI consumption while linking to depth to satisfy humans who want to linger. Measure beyond clicks: Track downstream conversions (print/pdfs, saves, subscriptions) because SGE can change where and how users engage. Avoid over-optimizing for zero-click: Provide the snapshot answer but make sure there are irresistible reasons to click through — unique tips, variations, printable print-friendly formats, or video demos.
Metaphorically: don't just memorize the recipe — be the chef who also tells why the sauce matters and how to rescue a runny béchamel. The AI will take the quick recipe, but readers will pay for nuance.
7. How to apply these lessons
Here's a practical checklist and roadmap you can apply to any recipe or how-to topic if your goal is snapshot visibility plus sustained engagement.
Prioritize the micro-answer
Build a concise, numbered "5-step" card at the top of the page. Keep each step under 20–30 words. Make it scannable for both users and AI extractors.
Implement robust structured data
Add full Recipe JSON-LD with step objects, cookTime, totalTime, recipeYield, ingredients array, images, and FAQ schema. Ensure the structured data exactly matches the visible content — avoid mismatches that confuse automated systems.
Demonstrate experience
Include an "tested X times" line, photos of intermediate steps, a video demo (with transcript), and an author bio noting relevant credentials or experience.
Optimize for speed and stability
Render the top-of-page card server-side, reserve image dimensions, compress assets, and aim for mobile LCP < 2s and CLS < 0.1. These are not vanity metrics; they affect whether your content is used in AI synthesis.
Craft an engaging "why click" proposition
Offer something behind the click: printable PDF, a video walkthrough, advanced tips, variants (vegetarian, gluten-free), or an ingredient swap chart. Make it easy to save or print.
Monitor and iterate with intent data
Use GSC to watch impressions and which queries trigger the snapshot, GA4 for engagement metrics, and server logs for click attribution. If snapshot impressions spike but clicks don't, optimize the reason-to-click elements.
Think in entities, not just keywords
SGE synthesizes knowledge; structure content around entities (lasagna, béchamel, ricotta) and relationships (step 1: preheat and prepare sauce) so the AI can accurately reference your content.
Protect your conversion funnel
Make essential interactions (print/save/email) accessible without walls. Avoid gating the core recipe behind email capture; instead, gate bonus content. That reduces friction for AI while preserving your growth channel.
Final thought, with a pinch of cynicism: the arrival of AI search didn't kill SEO; it changed the menu. If you keep serving long-winded intros and burying your steps like a secret family recipe in the basement, the AI will boil down your dish to a bland summary and serve somebody else’s appetizer. But if you plate a clean, authoritative five-step card and make the full-course experience irresistible, you win both the snapshot and the customer.
SGE snapshots are not a death knell; they're a new form of placement. Optimize like a chef who knows the front-of-house and the back-of-house — the AI will use your dish to make the restaurant look good, and diners will keep coming back when the main course is worth it.