There are a dozen JSON-LD types that technically apply to a blog post. Most are a waste of effort. Two of them consistently move AI citations and rich results within weeks. Here is the shortlist, ranked by yield against the hours it takes to ship, with the one we add on day one and the one that sounds smart on a conference slide but does nothing.
Schema is one of those areas where the gap between "technically correct" and "useful" is enormous. You can validate against schema.org all day and still ship JSON-LD that does nothing for your rankings, your rich results, or your AI citation frequency. We keep a shortlist, and the shortlist is shorter than most people expect.
Out of the forty plus types in the schema.org vocabulary, four cover almost every post we ever touch: Article, FAQPage, HowTo, and Product. Everything else is either rare enough to not matter in aggregate, or niche enough to belong to a specialist vertical (LocalBusiness for brick and mortar, Recipe for food sites, Event for venues, SoftwareApplication for a narrow slice of SaaS). The rest of this piece is about the four.
| Type | Effort | Citation yield | Verdict |
|---|---|---|---|
| FAQPage | Low | High | Ship first |
| HowTo | Medium | Medium-high | Ship when the post fits |
| Article | Low | Low-medium | Ship, but do not expect miracles |
| Product | Medium | Medium (for review posts) | Only on comparison or review pages |
FAQPage is the schema we add on the first pass, every time, no exceptions, on any post that can support even three plausible Q and A pairs. The reason is not mysterious. Three things are true about FAQPage that are not true about the others:
Speakable schema was introduced around 2019 to mark up passages intended for voice assistant read-out. On paper it sounds like the obvious GEO play. In practice it has never moved a metric we measure. No major LLM uses it as a citation signal. Google's voice surfaces are a rounding error in most libraries' traffic mix. The markup adds weight to every page it touches and returns nothing. We have never shipped Speakable on a client library and never had a client regret its absence.
If you read a conference deck that suggests Speakable is the future of voice search GEO, the deck is from 2020. Skip it.
HowTo is worth the effort when a post is genuinely procedural. A step by step tutorial with photos and numbered stages maps to HowTo naturally, and the schema earns citations in ChatGPT and Perplexity at a rate close to FAQPage. But it is a bad fit on an analytical essay, a comparison post, or a thesis piece. Do not bolt HowTo onto a post that is not really how-to; the engines notice, and over time they devalue your schema as a trust signal.
Article is the bare minimum for any editorial post. It earns no rich results of its own, but it is the signal that binds the post to an author, a publisher, a published date, and a main entity. LLMs use it to disambiguate. It is cheap, it is universal, and there is no reason not to ship it. It is the one schema we always include alongside FAQPage or HowTo; it is the base layer.
Product (with aggregateRating where honest ratings exist) belongs on review and comparison posts, full stop. If you are writing an unbiased comparison of five CRMs, the comparison block is a structured Product or Service array. If you are writing a thesis post on CRM market structure, it is not.
Below is a snippet generator. Pick your page type, fill in the title and URL, and it will emit a minimal, valid JSON-LD block you can paste into the head of the post. These are minimal by design: they are the bones, and you can extend them safely once the bones are in place.
The sequencing we follow on every Boost account: Article schema goes on every post in the first week, because it is cheap and sets the base layer. FAQPage goes on the top 20 posts in weeks two and three, because that is where the citation yield is. HowTo goes on any tutorial post that earns it. Product goes on review posts, only. Everything else is an optimization we may or may not get to, and it is never the reason a library starts compounding.
If you are running this yourself on a library of 40 posts, expect the full schema pass to take about four working days once you have the structural audit done. That is a lot less than most agencies quote for the same scope, and it is a lot more effective than buying a plugin that ships one generic Article block on every URL.