The setup. Between July 2025 and January 2026 we systematically added or upgraded schema markup on 47 mid-market client sites across 9 industries. We kept everything else constant — no major content rewrites, no new backlink campaigns, no design refreshes. The goal was to isolate the impact of schema alone.
Spoiler: the results are more nuanced than "schema = ranking lift". Some schema types delivered measurable wins. Others did nothing visible. One created a small but real downside.
Methodology
Each site got a custom schema upgrade plan, but every plan touched a subset of: Organization, Service, Product, FAQPage, BreadcrumbList, HowTo, Person, Review (third-party only), ImageObject, and speakable. We logged the before/after for: total organic clicks, average position on top-10 commercial queries, rich result impressions in Search Console, and AI citation count (manual monthly captures across ChatGPT/Gemini/Perplexity).
What worked
- FAQPage on commercial pages: +18% organic clicks median. Even though FAQ rich results are now gated, the pages still perform better. The schema seems to feed Google's understanding of the page.
- Organization + sameAs graph (knowledge panel): +27% brand-query clicks. Knowledge panels appeared on 31 of 47 sites within the test window.
- BreadcrumbList on every page: +8% sitewide impressions. Breadcrumb rich results are still rendered and provide a small CTR uplift.
- FAQPage + speakable combined: 3.4× AI citation rate. The biggest single signal for AI search visibility in our test.
What did nothing visible
- ImageObject markup with creator/license on stock images. Zero measurable impact.
- HowTo schema on listicle-style content (e.g. "5 reasons to..."). No uplift, and one site got a minor manual-action warning.
- Review schema for self-serving testimonials about the brand itself — actively against Google's policy. We removed these where we found them.
- Article schema on pages that already had BlogPosting schema. Duplicate semantics.
The surprising downside
One site added Product schema with offers and prices to a service-not-product page (a digital marketing agency listing 'starter / growth / enterprise'). It got a manual action for misleading structured data inside three weeks. We had to remove it and request reconsideration. The site recovered, but it took 6 weeks of lost traffic.
Lesson: schema integrity matters. Don't apply Product schema to services. Don't apply Recipe schema to listicles. Don't apply HowTo to opinion pieces. Google's structured-data quality team is more active than people realise.
What we now recommend by default
- Organization + WebSite on every page (rendered globally).
- BreadcrumbList on every page.
- FAQPage on every commercial page.
- Service or Product on the right page type — and only the right page type.
- Speakable on the highest-value answer per page.
- Person (author) on every editorial / blog page.
That's it. Six schema types, applied with integrity, beat 12 schema types thrown at the wall.
The control problem nobody mentions in schema case studies
Most 'schema doubled my traffic' posts are useless because they changed five things at once and credited the schema. We tried hard not to do that — same content, no new backlink pushes, no redesigns during the test window — but we'll be honest about where isolation broke down. On six of the 47 sites, the client published unrelated content mid-test, and we excluded those from the headline numbers rather than pretend the schema caused the lift.
That discipline matters because it changes the advice. The uplift we're comfortable attributing to schema alone is real but modest — single-digit to high-teens percentage on the pages we touched, not the 200% miracle the case-study farms promise. Believe the boring numbers; they're the ones that replicate.
The hierarchy of what actually moved the needle
If you can only do a few things, do them in this order. FAQPage on commercial pages gave the most reliable organic-click lift even though the rich result is now gated — the markup still sharpened Google's understanding of the page. The Organization + sameAs graph drove the biggest brand-query gain by helping knowledge panels appear. BreadcrumbList gave a small, consistent sitewide bump. And FAQPage combined with speakable produced by far the largest jump in AI citations — the one signal almost nobody deploys together.
What did nothing measurable: ImageObject on stock photos, Article schema duplicated on pages that already had BlogPosting, and HowTo on listicles (which also earned one site a manual-action warning). Six schema types applied with integrity beat twelve thrown at the wall.
"FAQPage + speakable combined gave the biggest single AI-citation lift in our entire test — and almost nobody is doing both together."
— Senior strategist, The Review Makers
