“Does schema markup help you get cited by ChatGPT and AI Overviews?”
Direct Answer
Schema markup does not rank you inside an AI answer. What it does is make your page cheaply parseable, so an engine can tell what the page is, who stands behind it, and which questions it answers without having to infer any of that from prose. In 2026 that matters more than it used to, because answer engines are summarising rather than linking. Five types carry almost all the weight for a small business: FAQPage, Article with a real author entity, Organization with sameAs, Service or Product with a priced Offer, and HowTo where the steps are genuinely sequential. Ship those five correctly and stop there.
What this guide covers
Part of our complete guide to search authority in 2026. If you have not read it yet, start with what AEO actually is — this page assumes you already know.
01What schema actually does for AI answers
There is a persistent myth that adding JSON-LD to a page makes ChatGPT or Google AI Overviews more likely to cite it. That is not how it works, and believing it leads people to bolt twelve schema types onto a thin page and wonder why nothing changed.
Schema is a parsing aid. It is a machine-readable statement of what a human would understand by reading the page: this is an article, this person wrote it, this is the organisation behind it, these are the questions it answers, this is what the service costs. An answer engine that has to infer all of that from prose will sometimes get it right and sometimes get it wrong. An engine reading clean markup gets it right every time, cheaply.
Why that matters more in 2026 than it did in 2023: answer engines are now the surface, not the index. AI Overviews appear on roughly 48% of Google searches globally by BrightEdge measurement, and when one appears, around 83% of those queries end with no click to any website at all. Separately, SparkToro measured the share of searches producing at least one click falling 9.51 percentage points between 2024 and 2026. Being the source a summary is built from is now a separate game from being the blue link someone clicks.
So the honest framing is this: schema is necessary, not sufficient. It will not rescue a page with nothing to say. It reliably stops a page that does have something to say from being skipped because the engine could not work out what it was.
02The five schema types that matter
| Type | What it resolves | Ship it when |
|---|---|---|
| FAQPage | Turns prose into explicit question-answer pairs an engine can lift verbatim | The page answers 3+ discrete questions |
| Article + author | Attaches a named, linkable human to the claims on the page | Every editorial page, without exception |
| Organization + sameAs | Ties the site to your real-world entity across LinkedIn, Crunchbase, directories | Once, sitewide. Most sites get this wrong or partial |
| Service / Product + Offer | States what you sell and what it costs, with a currency | Any commercial page. An Offer with no price is close to useless |
| HowTo | Marks genuinely sequential steps as steps | Only when the content is actually procedural. Do not force it |
Notice what is missing. BreadcrumbList, WebPage, ImageObject, SpeakableSpecification — all valid, all mostly noise for a small business trying to get cited. Your SEO plugin probably emits several of them already and that is fine. Adding more types is not the lever. Getting these five right is.
The author entity is the one people skip. An Article node whose author is a bare string (“admin”, or the site name) tells an engine nothing. An author node with a name, a URL to a real bio page, and sameAs links to a LinkedIn profile is a claim that can be verified. In a citation economy where engines are visibly biasing toward attributable sources, that gap is worth closing before you touch anything else.
03How to ship it without a plugin rewrite
You almost certainly do not need a new plugin. Most SEO plugins already output Article, WebPage, Organization and BreadcrumbList automatically. The usual real gaps are FAQPage, a properly linked author entity, and priced Offers on service pages.
Step 1 — audit what you already emit. Open any page, view source, and search for “application/ld+json”. Paste what you find into the Schema.org validator. You will usually find you have more than you thought, and that one or two nodes are half-populated.
Step 2 — add only what is missing, as one block. A single JSON-LD script in the page body is enough. It does not need to live in the head, and it does not need to be merged into the plugin graph. Engines read all JSON-LD blocks on the page.
Step 3 — mirror it in visible content. Every FAQ question in your markup must exist as visible text on the page. This is not a style preference, it is the documented requirement, and it is the single most common reason FAQ markup gets ignored.
Step 4 — fix Organization once, sitewide. One Organization node, correct legal name, logo URL, and sameAs pointing at every profile you actually control. Do this once and never think about it again.
04The mistakes that silently kill your markup
These are the four failures we see most often on audits, in rough order of how much damage they do.
Duplicate FAQPage nodes on one URL. Your plugin emits one, your page body emits another, and now the page declares two competing FAQ graphs. Engines commonly resolve this by ignoring both. If you add FAQ markup manually, confirm there is exactly one FAQPage node on the rendered page.
Markup describing content that is not on the page. Marking up reviews you do not display, prices you do not show, or questions that appear nowhere in the visible body is a policy violation, not a clever shortcut. It is also one of the few schema errors that can trigger a manual action.
Offers with no price. An Offer node with a currency and no numeric price communicates nothing. If you genuinely cannot publish a price, a priceRange or a “from” figure is far better than an empty Offer.
Stale dateModified. A page claiming it was updated in 2024 while the body text references this year is an internal contradiction, and it is trivially detectable. Keep dateModified honest or leave it out.
Related reading
If you are working through AI visibility systematically, these go in order:
- How to get cited in ChatGPT, Perplexity and AI Overviews — the content-side counterpart to this page.
- The GEO checklist for small business — ten steps, in sequence.
- How to track AI search traffic — because none of this is worth doing unmeasured.
- Seven fixes when ChatGPT will not cite your site — the diagnostic version.
If you would rather have someone run the audit and ship the fixes, that is what our AEO and GEO engagements are for.
05How to verify you are actually being cited
Validation is not verification. Validation tells you the syntax parses — use the Rich Results Test and the Schema.org validator, and watch Search Console enhancement reports for errors that appear after a deploy. Verification means asking the engines your actual target questions, in a clean session, and recording whether your domain shows up in the citation list. Do that monthly for a fixed list of ten questions and you will have a real trend line inside a quarter. Nothing else in AEO gives you a cleaner read.
?FAQs
Does schema markup directly improve AI citations?
Which schema types matter most for AEO?
Do I need a plugin to add schema markup?
How do I know if my markup is working?
Can bad schema hurt me?
Want this built for your business?
We win answer engines and snippets. Book a free 30-minute audit and we will map it to your funnel.
Explore AEO Services →Book a free audit →