Next SEO

What Next SEO is today

Next SEO (github.com/garmeeh/next-seo) is a TypeScript library, MIT-licensed and maintained since 2018, that gives a Next.js project ready-made components for JSON-LD structured data. You import a component, pass it typed props, and it renders the script tag that tells a search engine what the page is: an article with an author and a date, a product with a price and a return policy, a set of FAQs, a breadcrumb trail, an organization, a video, a job posting. The README documents more than two dozen of them, and a JsonLdScript primitive plus a library of input processors for the Schema.org types it does not cover.

The more interesting fact about the project is what it stopped doing. Next SEO began as the standard way to manage title and meta tags in Next.js, back when the framework had no opinion about them. Since the App Router shipped a first-class metadata API, the framework does that job natively — and rather than compete, the library's README now tells you plainly to use the built-in function for standard meta tags and keeps its own surface on structured data, with the older Pages Router components moved behind a separate import. A dependency that gets smaller when the platform absorbs part of its job is a healthier signal than one that keeps growing.

Where structured data actually helps

JSON-LD is a block of machine-readable claims about a page, sitting alongside the human copy. It is not a secret channel and it does not make a weak page strong. What it changes is how legible your page is to systems that have to decide, without reading it the way a person does, what the page is and whether to reuse it.

• Eligibility for rich results Some types can change how a listing looks in search — reviews, FAQs, products, events, recipes. Correct markup makes a page eligible; it never obliges a search engine to show anything.
• Legibility for answer engines Explicit author, date, entity and question-answer pairs give a retrieval system something unambiguous to lift, which is the mechanical half of generative engine optimization.
• Internal consistency Filling in a typed component forces you to state who wrote a page and when, which is exactly where hand-built sites quietly disagree with themselves.

For a small SaaS or store the useful set is short: an organization or software-application type on the site, article and FAQ types on content, breadcrumbs for structure, and product and review types if you sell physical goods. The library also accepts a nonce on every script it renders, which matters if you run a strict content-security policy — the kind of detail that costs an afternoon when you hand-roll it. Background on the concept itself is in schema markup.

What it does not do

Structured data is not a ranking lever. Google has been consistent that valid markup makes a page eligible for certain presentations rather than ranking it higher, and no amount of it compensates for a page nobody had a reason to open. Treat it as the difference between a page that is understood and a page that has to be guessed at.

Which puts a library like this firmly on the build side of the build-then-grow split. A correctly marked-up page still needs a question worth answering, a reason for someone to link to it, and enough of them to add up to a topic. If you are choosing what to publish rather than how to annotate it, SEO for a brand new website is the upstream decision, and open-seo is the research side of the same job — this is the rendering side.

A first-party note on shipping schema

At AgentCeres — the AI growth team at agentceres.com — we write our JSON-LD by hand in the App Router rather than through a library, so this is a comparison from the other side of the choice. The mistake we made was not syntactic. Our question-and-answer pages originally emitted the QAPage type, which reads like the obvious fit, and it was wrong: that type describes pages where visitors post competing answers, and ours are written by us. We changed them to FAQPage. Nothing about the page changed except the truthfulness of the claim the markup was making — and a library would have rendered the wrong type just as cleanly as the right one.

The second lesson cost more. Markup is a set of claims, and claims go stale silently, because nothing on the page looks broken when they do. Every page in our own /ask, /glossary and /grow sections is a typed entry in source control with a test suite that fails the build on a dangling internal link, an over-long meta title or a missing publish date, so the values feeding the schema cannot drift without someone noticing. Next SEO buys you correct syntax and saves real time; nothing buys you correct facts except deciding which of them are allowed to be typed by hand. That distinction is most of what our SEO content specialist actually enforces.