# AI Generation & FAQ

Egg Blocks are designed to be easy to generate with AI. Instead of writing each block by hand, you can ask ChatGPT, Claude, or any AI chat to draft a complete article — intro, key takeaways, comparison table, pros/cons, FAQ, conclusion, and the rest — and paste the result straight into WordPress.

This page walks through the workflow, points out the gotchas, and answers common questions.

### Contents

1. [What you get](#what-you-get)
2. [Quickstart](#quickstart)
3. [Which AI chat to use](#which-ai-chat-to-use)
4. [How to prompt](#how-to-prompt)
5. [Working with product blocks](#working-with-product-blocks)
6. [Tips for better output](#tips-for-better-output)
7. [FAQ / troubleshooting](https://ce-docs.keywordrush.com/faq)

***

### What you get

The **EggBlocks Writer** is a small instruction file (a "skill") that teaches any modern AI chat how to write Egg Blocks articles correctly. Once the AI has read it, you can ask for:

* **A complete article** — e.g. "a buying guide for standing desks under $500"
* **A single section** — e.g. "just the pros-cons and FAQ for this product"
* **A single block** — e.g. "one callout warning about counterfeit chargers"

The AI outputs ready-to-paste **Gutenberg markup**. You copy it, paste it into the Code Editor view in WordPress, switch back to the Visual Editor, and your article is there — fully structured with intros, takeaways, tables, FAQs, and everything else.

What the AI handles:

* Picking appropriate blocks for the article type
* Filling in titles, body text, variants, and TOC metadata
* Following the composition patterns from the Introduction recipes
* Generating coherent, consistent voice across the whole article

What you still handle:

* Assigning real Content Egg products to product blocks
* Fact-checking claims (AIs can invent specs or prices)
* Editorial polish and final voice

***

### Quickstart

#### 1. Load the skill

Paste this URL into ChatGPT, Claude, or any AI chat with web-fetch capability:

```
https://www.keywordrush.com/skills/eggblocks-writer/SKILL.md
```

Ask the AI to read it. You should see a short greeting confirming the skill loaded.

<figure><img src="https://4254262503-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-M3fnB7iKYwDc1Xhr3H3%2Fuploads%2FrNvc4b8VA7UstKiBwyy7%2Fimage.png?alt=media&#x26;token=03134c69-63c7-41a7-b73b-76a3a8df376b" alt="" width="563"><figcaption></figcaption></figure>

If you don't see a confirmation, the AI likely didn't fetch the URL. Try a different chat that has web access, or paste the SKILL.md contents directly into the conversation.

#### 2. Ask for what you need

```
/blocks single-product review of Sony WH-1000XM5
```

Or, without the slash command:

```
write a roundup of 5 ergonomic chairs as blocks
```

Or just name a block type:

```
faq block for best mechanical keyboards
```

<figure><img src="https://4254262503-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-M3fnB7iKYwDc1Xhr3H3%2Fuploads%2FrrO6l08pAEELsEsff5Sb%2Fimage.png?alt=media&#x26;token=39c942d0-ace4-4428-88d1-cd5d41834ad6" alt="" width="563"><figcaption></figcaption></figure>

#### 3. Paste into WordPress

1. Copy the fenced code block from the AI's response.
2. Open your post in the WordPress block editor.
3. Paste the markup.

<figure><img src="https://4254262503-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-M3fnB7iKYwDc1Xhr3H3%2Fuploads%2Fi4diE5HwT7ABBQ1HQfCT%2Fimage.png?alt=media&#x26;token=13698491-9db7-435e-b99e-ad4dab35ef50" alt="" width="563"><figcaption></figcaption></figure>

#### 4. Assign products

If you already have Content Egg products on the page, use the **Copy all product references** button in the metabox toolbar before prompting the AI. It copies a JSON array of every product — title, URL, module ID, and unique ID — to your clipboard. Paste that data into the chat together with your article request and the AI will fill in `product_ref` automatically for every block it can match.

Without pasted product data, product-bound blocks (product-card, quick-picks, comparison-table, where-to-buy) are left with empty product slots. Click each one in the block editor, open the right sidebar, and use **Select product** to bind the matching Content Egg product. The AI output will list which blocks still need manual assignment at the end.

<figure><img src="https://4254262503-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-M3fnB7iKYwDc1Xhr3H3%2Fuploads%2FGGmyhlDQF3x21PuxCN1n%2Fimage.png?alt=media&#x26;token=d3844187-4ca2-4cad-9d84-1865f3e15640" alt="" width="563"><figcaption></figcaption></figure>

***

### Which AI chat to use

The skill works with any AI chat that can:

* Read a URL when asked
* Follow structured instructions
* Produce enough output to cover a full article

Confirmed-working chats (as of writing):

* **Claude** (claude.ai) — reads the URL reliably, strong adherence to the schema, best for long articles.
* **ChatGPT** (chat.openai.com) — works with web browsing enabled. Use GPT-4-class models; older/smaller models tend to invent block types.
* **Google Gemini** — works when web access is enabled.
* **Any Claude/GPT-powered assistant** (Perplexity, Poe, Copilot Chat, custom agents) — works if the underlying model is capable enough.

Smaller or older models often struggle. If the output looks wrong (wrong block names, invented fields, missing markup), try a stronger model before blaming the skill.

***

### How to prompt

The skill operates in two modes:

#### Conversation mode (default)

Just chat. Plan the article, discuss products, refine the outline, pick a tone. No block markup is produced. Use this to agree on scope before generating content.

Example:

> *"I want to write about portable monitors under $300. Can you help me plan the article?"*

The AI will respond in plain prose — asking questions, suggesting an outline, recommending blocks.

#### Block mode (triggered)

Block mode produces the pasteable markup. Trigger it with any of:

* **Slash command** — start with `/blocks`
* **Phrase** — include *"as blocks"* anywhere in the message
* **Naming a block type** — e.g. *"pros-cons for X"*, *"faq for Y"*

Example:

> *"/blocks single-product review of the Logitech MX Master 3S"*

Follow-ups stay in block mode automatically. So after the AI generates an article, you can say:

> *"make the intro shorter"*
>
> *"replace pick 3 with the Razer Pro Click"*
>
> *"add a callout warning about wrist strain"*

…and the AI will regenerate the affected blocks.

#### Good prompt examples

| Prompt                                                    | What you'll get                                  |
| --------------------------------------------------------- | ------------------------------------------------ |
| `Plan a buying guide for mechanical keyboards under $150` | Conversation: outline and suggestions, no markup |
| `/blocks single-product review of Sony WH-1000XM5`        | Full review article, Gutenberg markup            |
| `pros-cons block for BEAVERLAB Finder 4.0`                | A single pros-cons block                         |
| `as blocks: roundup of 5 ergonomic chairs`                | Full roundup article                             |
| `no product blocks — explainer on USB-C standards`        | Full article, no commerce blocks                 |

#### Scope keywords

* **Full article** is the default when you ask for a "review", "guide", "roundup", etc.
* **Section only** — say *"just the conclusion section"* or *"only the comparison-table and verdict"*.
* **Block only** — name a single block type, e.g. *"one callout about humidity"*.
* **No commerce** — include *"no product blocks"*, *"text only"*, or *"prose only"* to skip all product-bound blocks.

***

### Working with product blocks

When you mention specific products by name (e.g. "Sony WH-1000XM5", "Herman Miller Aeron"), the AI includes product-bound blocks automatically: product-card, quick-picks, comparison-table, where-to-buy, and so on.

#### Shortcut: copy products from the metabox

If the products are already saved on the page in Content Egg, click the **Copy all product references** button in the metabox toolbar. It copies a JSON array to your clipboard in this format:

```
[
  { "title": "Sony WH-1000XM5", "orig_url": "https://...", "product_ref": { "module_id": "Amazon", "unique_id": "us-B09MGVFLRR" } },
  ...
]
```

<figure><img src="https://4254262503-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-M3fnB7iKYwDc1Xhr3H3%2Fuploads%2FlLJIyIDDwkt4WPHml7MK%2Fimage.png?alt=media&#x26;token=48cb5530-e5fd-41b8-b731-fc3328ddda0f" alt="" width="563"><figcaption></figcaption></figure>

Paste that array into your AI prompt alongside the article request. The AI reads it and assigns `product_ref` automatically to every block it can match by product title. Only blocks it cannot match are left empty.

Without pasted product data, all product slots are left empty. You assign real products in WordPress:

1. Paste the article into the block editor.
2. Click a product-bound block (e.g. a product-card with "Unassigned product" in it).
3. In the right sidebar, click **Select product**.

The AI will list which blocks still need manual assignment at the end of its output — something like:

> "Assign products to: product-card (Logitech MX Master 3S), quick-picks row 1, row 2, row 3, comparison-table rows 1–3, where-to-buy (MX Master 3S)."

If you want an article with no product blocks at all, say so in your prompt ("no product blocks", "prose only", "explainer — no commerce"). The AI will still write reviews or guides, but in prose form without any product-bound blocks.

***

### Tips for better output

#### Plan first, generate second

For anything longer than a single block, ask the AI to propose an outline before generating the article. It's cheaper to tweak an outline than a 2,000-word draft.

> *"Plan a buying guide for portable monitors — don't generate blocks yet."*

Once the outline looks right, say *"go ahead and generate"*.

#### Name specific products

Vague prompts produce vague content. Instead of *"best headphones"*, say *"best headphones — Sony WH-1000XM5, Bose QC Ultra, Apple AirPods Max, Sennheiser Momentum 4, Anker Space One"*. The AI will use those exact products in product blocks and write around them.

#### Keep revisions small

Don't say *"rewrite the whole article with a different tone"*. Do say *"make the intro more conversational"* or *"shorten pick 2's description"*. Targeted revisions produce better results and waste fewer AI tokens.

#### Fact-check everything

AIs invent specs, prices, release dates, and feature lists. Treat every factual claim as a draft. Before publishing, verify:

* Product specifications
* Prices (these come from Content Egg automatically once products are assigned, but editorial mentions won't)
* Launch dates or version numbers
* Quoted review scores

#### Regenerate if structure is wrong

If the AI outputs something that doesn't look like valid markup, or mixes block types incorrectly, just say *"that's invalid, try again"*. The skill tells it to self-correct. If multiple attempts fail, you're probably using a weaker model — switch to a stronger one.

***

### FAQ / troubleshooting

#### The AI writes plain paragraphs instead of Egg Blocks

You're in conversation mode. Trigger block mode: prefix your message with `/blocks`, include *"as blocks"*, or name a specific block type.

#### The AI says it doesn't know what Egg Blocks are

It didn't read the skill URL. Ask it again: *"please fetch and read <https://www.keywordrush.com/skills/eggblocks-writer/SKILL.md>"*. If it still won't fetch URLs, copy the contents of that URL and paste them directly into the chat.

#### The AI invented block types that don't exist

Usually a weaker-model issue. Try a more capable model (Claude Sonnet/Opus, GPT-4+, Gemini Pro). If the problem persists, ask the AI to re-read the skill and list the 21 available blocks before generating.

#### I pasted the markup but some blocks look broken

Most "broken" blocks are caused by one of:

* **Truncated output** — the AI stopped mid-article. Ask it to continue.
* **A single missing product** — unassigned product slots render as empty cards. Assign products.
* **Wrong variant** — switch the variant in the sidebar.
* **Copy-paste artifacts** — you missed the opening or closing of the fenced code block. Recopy and repaste.

#### Can I generate content in a language other than English?

Yes. Ask in your target language, or add *"in German"* / *"write in Spanish"* to the prompt. The block schema itself is language-neutral — Egg Blocks render whatever text the AI produces.

#### Can I use this with a local LLM (Ollama, llama.cpp)?

Only if the model is capable enough to follow the schema and produce long structured output. Most 7B–13B local models aren't. Try a 70B+ model, or use the hosted options.

#### Does the AI have access to my Content Egg products?

No. The AI never sees your Content Egg data. It only produces editorial structure and leaves product slots empty for you to fill in WordPress. Your product catalog stays private.

#### Can the AI set prices or stock status?

No — and it shouldn't. Prices, stock, buy URLs, and merchant logos all come from Content Egg automatically once you assign a product. That's the whole point of product binding: prices update on a schedule so your articles never carry stale numbers.

#### What if the skill changes?

The skill is a plain text file at the URL. New versions of Egg Blocks may bring new block types or new variants. The skill is updated to match. If your AI is giving outdated output, ask it to re-fetch the URL in a fresh chat.

#### Where do I report issues with the skill?

If you find bugs, cases where the AI produces invalid markup consistently, or features you'd like supported, [contact Content Egg support](https://www.keywordrush.com/contact) with the exact prompt, the AI's response, and the model you used.


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://ce-docs.keywordrush.com/egg-blocks/ai-generation-and-faq.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.