AI SEO content generator: what good output looks like, and how to check it

It takes a topic, usually a query you want to rank for, and returns an article built for search. The article is only part of it. A post also needs a title, an SEO title that fits in a results page, a meta description, a short summary, a URL slug, and usually a category and an image.

Most generators stop at the text. You then paste it into your CMS, write the fields yourself, find images, and publish. On WordPress that is tedious. On a headless CMS like Contentful, where every part of a post is its own field and the body is structured rich text, it is most of the work.

These are the parts of a post and the specification Frogpost writes to. Use it as a checklist for any generator you are considering.

Frogpost also writes in the second person, in a conversational tone, with no placeholders, no bracketed notes, and no filler lines like "in today's fast-paced world".

A model asked to write about "invoice reminders" for a company it knows nothing about writes what a company like that usually says. The advice is plausible, the product is described vaguely, and sometimes a feature appears that does not exist. Readers notice, and it is the part that does real damage: a post that promises something your product cannot do.

Give it your real product facts

When you connect a project, Frogpost reads your website first. It fetches up to twelve pages, starting with the ones that describe the product: documentation, features, how it works, pricing, integrations, about and FAQ. It skips the pages that do not, such as the blog, legal pages, careers, sign in and checkout.

It then writes a briefing of under 600 words with five parts: what the business does, who it is for, what it offers using the product's own names, the vocabulary the business uses and avoids, and concrete proof points like integrations and pricing. The briefing is told to state only what the pages say, and to say so plainly if the site is thin.

Add your own words

You also get a short description of the business, drafted from that briefing for you to correct. It sets the voice and the audience. The crawled briefing supplies the facts. Every post is written with both, and the writer is told never to claim a capability the briefing does not list.

For a topic like "invoice reminder schedule" on an invoicing product, the result arrives as structured fields, not a single block of text. Each field goes into its own place in your CMS.

{
  "title": "How to Send Invoice Reminders Without Annoying Clients",
  "slug": "invoice-reminder-schedule",
  "excerpt": "Late payments are rarely about unwillingness. Most clients simply lose track...",
  "category": "Getting paid",
  "metaTitle": "Invoice Reminder Schedule: When and How to Follow Up",
  "metaDescription": "A reminder schedule that gets invoices paid without straining the relationship, with timing, wording and when to automate each follow up.",
  "body": "## Why invoices get paid late\n\n### The invoice never reached the right person\n..."
}

The body is written in markdown and converted to Contentful rich text when the entry is created, so headings, lists, bold text, links and images arrive as real rich text nodes that your renderer already handles.

A generator cannot create demand. A well written post on a topic nobody searches publishes on time and gets no visits. Before you queue a topic, check it in Google Keyword Planner, which is free, and look at who ranks for it today. A good topic is a question your customers actually ask, phrased the way they would type it.

Frogpost can suggest topics for a project. Suggestions are filtered against everything that project has already published or queued, so you do not end up with three posts on the same question. Treat them as a starting list, and check the ones you pick.

1. Read the first two paragraphs. They should answer the query directly, not warm up to it.
2. Check every sentence that describes your product against your own pages.
3. Check the SEO title and meta description in your CMS. They are what people see in search.
4. Open every link. Links to your site should go to the right page, and links to earlier posts should fit the sentence.
5. Look at the images on your live site, not only in the CMS.
6. Check the category. It should be one of the values your blog already uses.
7. Ask whether you would send this post to a customer who asked the question.

For your first few posts, save them as drafts and do this for each one. Once they consistently pass, a schedule is safe.

The last step is the one most generators leave to you. On Contentful, Frogpost creates the entry in your own blog content type and fills the fields it recognises: title, slug, excerpt, rich text body, SEO title, meta description, publish date and featured image. Images are uploaded to your space as assets and published. Categories and tags are chosen from the values you allow. Earlier posts on the same blog are linked where they fit.

You choose how it goes out: as a draft to review, published at once, or from a queue at the hours you set. Your site picks up new entries the way it picks up any other publish, including any webhook you already have for rebuilding or revalidating the blog.

The last row is the honest limit. A generator can explain and structure what is known about a topic and your product. It cannot supply a customer story, original data, or a strong opinion you have not written down. The best blogs mix both: generated posts for the steady, searchable questions, and your own posts for what only you can say.

• Pages built from data, like one page per integration or per city. Those need a template and a dataset.
• News, announcements and anything that depends on something that happened this week.
• Posts that rest on your original research, numbers or customer stories.
• Topics nobody searches. No amount of writing fixes missing demand.