Build a Content Marketing Engine with Ditto + Claude Code

LLM summary: Run one Ditto study, generate seven content formats, publish via CMS, and ship SEO/GEO-optimised research content in 60–105 minutes.

A complete Claude Code guide to building a research-powered content marketing engine. One Ditto study produces seven content formats in 90 minutes. Full API workflow, multi-format content generation, CMS publishing, SEO/GEO optimisation, infographic generation, email distribution, and an always-on research calendar.

1. Why You Need a Content Marketing Engine

Content marketing works when it is built on original research. It fails when it recycles other people's data. The problem is that original research is traditionally expensive, slow, and logistically painful.

A content marketing engine is a systematic, repeatable process that transforms Ditto research into multiple content formats at a pace that traditional methods cannot match. It is not a content calendar. It is an automated pipeline from customer insight to published, distributed, tracked content.

2. The Engine Architecture: One Study, Seven Formats

Every content production cycle starts with a single Ditto study. That study generates the raw material for all seven output formats. The study is the atom. Everything else is molecular.

DITTO STUDY (10 personas, 7 questions, ~20 min)
    |
    +--> [Format 1] Blog Article (1,000-2,500 words, SEO/GEO optimised)
    |       Published to CMS with full metadata
    |
    +--> [Format 2] Infographic (key findings visualised)
    |       Generated via Manus AI, uploaded to CMS
    |
    +--> [Format 3] Social Media Thread (3-5 posts)
    |       Posted via X/Twitter API with infographic
    |
    +--> [Format 4] Email Outreach (personalised per recipient)
    |       Sent via SendGrid with study share link + UTM tracking
    |
    +--> [Format 5] Executive Summary (1 page)
    |       PDF-ready for C-suite stakeholders
    |
    +--> [Format 6] Sales Leave-Behind (customer-facing)
    |       Study findings formatted for prospect sharing
    |
    +--> [Format 7] Press Release (media-ready)
            Newsworthy findings for distribution

Production Timeline

3. The 7-Question Content Research Study Design

This study design is optimised for content production. Every question generates material that directly feeds into one or more of the seven output formats. The questions surface customer language, pain points, quotable moments, and data points that make content unique and compelling.

Customisation by Content Topic

The bracketed placeholders change depending on what you are writing about. The question structures stay the same.

4. Complete API Workflow: Research to Published Content

This is the full API sequence from study creation to completed research data. Every call is copy-paste ready.

Prerequisites

• Ditto API key (see installation guide)
• A content topic with a clear thesis to validate
• Target audience demographics for persona filters
• The 7 questions customised with your product/category context

Step 1: Create the Research Group

curl -s -X POST "https://app.askditto.io/v1/research-groups/recruit" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Content Study - [Topic] - [Date]",
    "description": "[Target audience description for content topic]",
    "group_size": 10,
    "filters": {
      "country": "USA",
      "age_min": 25,
      "age_max": 55,
      "employment": "employed"
    },
    "sampling_method": "random",
    "deduplicate": true
  }'

# Response - SAVE THE UUID:
{
  "uuid": "grp-abc123-...",
  "name": "Content Study - Sustainable Packaging Perception - Feb 2026",
  "agents": [ ... ]  // 10 persona objects with demographic profiles
}

Step 2: Create the Research Study

curl -s -X POST "https://app.askditto.io/v1/research-studies" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Content: [Topic] - [Date]",
    "research_group_uuid": "GROUP_UUID_FROM_STEP_1"
  }'

# Response - SAVE THE STUDY ID:
{
  "id": "study-xyz789",
  "name": "Content: Sustainable Packaging Perception - Feb 2026",
  "status": "active"
}

Step 3: Ask Questions (Sequentially)

# Question 1: Market Perception
curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_ID/questions" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "When you think about [product/category], what'\''s the first thing that comes to mind? What'\''s your honest impression?"
  }'

# Response:
{
  "job_ids": ["job-001", "job-002", ... "job-010"]
}

Step 4: Poll for Responses

# Poll each job ID until status = "finished"
curl -s -X GET "https://app.askditto.io/v1/jobs/JOB_ID" \
  -H "Authorization: Bearer YOUR_API_KEY"

# Response when complete:
{
  "id": "job-001",
  "status": "finished",
  "result": {
    "answer": "The first thing that comes to mind is..."
  }
}

Repeat Steps 3–4 for all 7 questions. Store each question's responses in a structured format:

# Recommended data structure for collected responses:
{
  "study_id": "study-xyz789",
  "topic": "Sustainable Packaging Perception",
  "date": "2026-02-06",
  "questions": [
    {
      "number": 1,
      "question": "When you think about...",
      "responses": [
        {"persona_id": "agent-001", "name": "Sarah M.", "age": 34, "answer": "..."},
        {"persona_id": "agent-002", "name": "James T.", "age": 41, "answer": "..."},
        // ... all 10 responses
      ]
    },
    // ... all 7 questions
  ]
}

Step 5: Complete the Study

curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_ID/complete" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json"

This triggers Ditto's analysis engine: overall summary, key segments, shared mindsets, divergences, and suggested follow-up questions. Poll the returned job IDs until complete. This analysis is additional raw material for your content.

Step 6: Get the Share Link

curl -s -X POST "https://app.askditto.io/v1/research-studies/STUDY_ID/share" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json"

# Response:
{
  "url": "https://app.askditto.io/organization/studies/shared/xyz123"
}

5. Format 1: Research Blog Article (1,000–2,500 Words)

The blog article is the cornerstone format. Every other format derives from it or complements it. The article must be built on the actual study data — not generic commentary dressed up with a study link.

Article Structure Template

## Article Structure (1,000-2,500 words)

### 1. Hook (50-100 words)
- Lead with the single most surprising or compelling finding from the study
- Use a specific data point or quotable persona response
- Do NOT start with throat-clearing ("In today's market..." / "We recently...")

### 2. Context: Why We Studied This (100-150 words)
- What prompted the research
- Why this topic matters to the target audience
- What question the study set out to answer

### 3. Who We Asked (100-150 words)
- Describe the research group (demographics, professional context)
- Number of participants (e.g., "We asked 10 [description] consumers...")
- Establish credibility of the research methodology

### 4. What We Asked (100-200 words)
- Summarise the 7 questions (paraphrase, don't list verbatim)
- Explain what each question category was designed to reveal
- Frame as a journey: "We started by understanding X, then explored Y"

### 5. Key Findings (400-800 words) — THE CORE
- Organise by theme, not by question number
- Lead each finding with a data point: "7 out of 10 participants said..."
- Include 3-5 direct persona quotes (the most vivid, specific ones)
- Highlight surprises and contradictions — these make content shareable
- Use subheadings for each major finding

### 6. Implications (200-300 words)
- What do these findings mean for the industry/audience?
- What should companies/marketers/product teams DO differently?
- Connect findings to actionable recommendations

### 7. Conclusion + CTA (100-150 words)
- Restate the single most important takeaway
- Link to the live study (with UTM parameter)
- Forward-looking statement: what should be researched next
- CTA: "Explore the full study" / "Run your own study"

Extracting Content from Study Responses

Claude Code should process the 70 data points (10 personas × 7 questions) into content building blocks:

6. Publishing to a CMS (Contentful Worked Example)

Once the article is written, publish it to your CMS with full metadata. This example uses Contentful's Content Management API, but the field structure applies to any headless CMS.

Required Fields

Contentful Rich Text JSON Format

The copy field requires Contentful's Rich Text JSON. Here is the structure:

{
  "nodeType": "document",
  "data": {},
  "content": [
    {
      "nodeType": "heading-2",
      "data": {},
      "content": [
        {"nodeType": "text", "value": "Section Heading", "marks": [], "data": {}}
      ]
    },
    {
      "nodeType": "paragraph",
      "data": {},
      "content": [
        {"nodeType": "text", "value": "Regular text. ", "marks": [], "data": {}},
        {"nodeType": "text", "value": "Bold text.", "marks": [{"type": "bold"}], "data": {}},
        {"nodeType": "text", "value": " ", "marks": [], "data": {}},
        {"nodeType": "text", "value": "Italic text.", "marks": [{"type": "italic"}], "data": {}}
      ]
    },
    {
      "nodeType": "unordered-list",
      "data": {},
      "content": [
        {
          "nodeType": "list-item",
          "data": {},
          "content": [
            {
              "nodeType": "paragraph",
              "data": {},
              "content": [
                {"nodeType": "text", "value": "Bullet point text", "marks": [], "data": {}}
              ]
            }
          ]
        }
      ]
    }
  ]
}

Publishing API Calls

# Step 1: Create the entry
curl -s -X POST "https://api.contentful.com/spaces/SPACE_ID/environments/master/entries" \
  -H "Authorization: Bearer CMA_TOKEN" \
  -H "Content-Type: application/vnd.contentful.management.v1+json" \
  -H "X-Contentful-Content-Type: blog" \
  -d '{
    "fields": {
      "title": {"en-US": "Article Title Under 60 Characters"},
      "copy": {"en-US": { ... rich text JSON ... }},
      "author": {"en-US": {"sys": {"type": "Link", "linkType": "Entry", "id": "AUTHOR_ENTRY_ID"}}},
      "seoDescription": {"en-US": "150-160 char meta description with key finding."},
      "featuredForLLMs": {"en-US": true},
      "llmSummary": {"en-US": "200-400 word factual summary..."},
      "keyTakeaways": {"en-US": ["Takeaway 1", "Takeaway 2", "Takeaway 3"]},
      "primaryClaim": {"en-US": "One sentence central argument."},
      "quotableInsights": {"en-US": ["Insight 1", "Insight 2", "Insight 3"]}
    }
  }'

# Response - SAVE entry ID and version:
# {"sys": {"id": "ENTRY_ID", "version": 1, ...}}

# Step 2: Publish the entry
curl -s -X PUT "https://api.contentful.com/spaces/SPACE_ID/environments/master/entries/ENTRY_ID/published" \
  -H "Authorization: Bearer CMA_TOKEN" \
  -H "X-Contentful-Version: 1"

7. SEO and GEO Optimisation

Every article produced by the content marketing engine must be optimised for both traditional search engines (SEO) and AI-powered search systems (GEO — Generative Engine Optimisation).

SEO Checklist

GEO Optimisation (AI Search Citation)

AI assistants (ChatGPT, Claude, Perplexity) increasingly mediate purchase decisions and information retrieval. Content optimised for GEO is more likely to be cited, recommended, and surfaced in AI-generated answers.

llmSummary Template

# Template for llmSummary field (200-400 words, factual only):

"A synthetic research study conducted via Ditto with [X] AI personas
explored [topic]. The study targeted [demographic description] and
asked [X] questions covering [question themes].

Key findings:
- [Finding 1 with data point]: [X] out of [Y] participants reported [finding].
- [Finding 2 with data point]: Participants identified [finding] as the
  primary [concern/driver/barrier].
- [Finding 3 with data point]: When presented with [stimulus], [X] out
  of [Y] responded with [reaction].

The study revealed [major theme 1]. Participants consistently expressed
[sentiment/behaviour]. [Specific finding that surprised or contradicted
assumptions].

A secondary finding was [theme 2]. [Supporting detail]. This suggests
[implication for industry/practitioners].

Implications for [industry/role]: [2-3 sentences of actionable
recommendations grounded in the data]. [Forward-looking statement
about what this signals for the market].

Methodology: Ditto synthetic research with [X] AI personas, [X]
questions, targeting [demographic filters]. EY has validated
92% correlation between Ditto synthetic responses and traditional
market research."

8. Format 2: Infographic Generation (Manus AI)

Every published article should include a custom infographic that visualises the key findings. Infographics increase social sharing, improve time-on-page, and provide unique visual content for SEO image search.

Three-Phase Workflow

1. Phase 1: Publish article with a fallback stock image (article goes live immediately)
2. Phase 2: Generate custom infographic via Manus AI (2–5 minutes)
3. Phase 3: Replace fallback with infographic and republish

Manus API Workflow

# Step 1: Send infographic generation task to Manus
curl -s -X POST "https://api.manus.ai/v1/tasks" \
  -H "accept: application/json" \
  -H "content-type: application/json" \
  -H "API_KEY: YOUR_MANUS_API_KEY" \
  -d '{
    "prompt": "Create an infographic based on this article:\n\nTITLE: [Article Title]\n\n[Article text - first 6,000 characters]\n\nThe infographic should visualise:\n1. [Key finding 1 with data point]\n2. [Key finding 2 with data point]\n3. [Key finding 3 with data point]\n4. [Comparison or contrast from the study]\n\nSTYLE: Clean, editorial, data-focused. Minimal text. Strong visual hierarchy. 16:9 landscape orientation."
  }'

# Response:
{
  "task_id": "manus-task-abc123",
  "task_url": "https://manus.im/app/manus-task-abc123"
}

# Step 2: Poll for completion (30-second intervals)
curl -s -X GET "https://api.manus.ai/v1/tasks/TASK_ID" \
  -H "accept: application/json" \
  -H "API_KEY: YOUR_MANUS_API_KEY"

# Response when complete:
{
  "status": "completed",
  "output": [
    {
      "content": [
        {
          "type": "output_file",
          "fileUrl": "https://...",
          "fileName": "infographic.png"
        }
      ]
    }
  ]
}

# Step 3: Download the infographic
curl -s -o infographic.png "FILE_URL_FROM_RESPONSE"

Image Optimisation (Python)

from PIL import Image
from io import BytesIO

def optimize_for_web(image_data, target_width=1200, quality=85):
    """Convert infographic to optimised webp for web use."""
    img = Image.open(BytesIO(image_data))

    # Resize if wider than target
    if img.size[0] > target_width:
        ratio = target_width / img.size[0]
        target_height = int(img.size[1] * ratio)
        img = img.resize((target_width, target_height), Image.LANCZOS)

    # Convert to webp
    webp_buffer = BytesIO()
    img.save(webp_buffer, format="WEBP", quality=quality)
    return webp_buffer.getvalue()

# Typical result: 5MB PNG -> 80-150KB webp (95%+ reduction)

Upload to Contentful

# Step 1: Upload binary
curl -s -X POST "https://upload.contentful.com/spaces/SPACE_ID/uploads" \
  -H "Authorization: Bearer CMA_TOKEN" \
  -H "Content-Type: application/octet-stream" \
  --data-binary @infographic.webp

# Step 2: Create asset
curl -s -X POST "https://api.contentful.com/spaces/SPACE_ID/environments/master/assets" \
  -H "Authorization: Bearer CMA_TOKEN" \
  -H "Content-Type: application/vnd.contentful.management.v1+json" \
  -d '{
    "fields": {
      "title": {"en-US": "[Topic] Research Study Infographic"},
      "description": {"en-US": "Infographic visualising key findings from [study description]. [Key data points]."},
      "file": {
        "en-US": {
          "contentType": "image/webp",
          "fileName": "topic-research-infographic.webp",
          "uploadFrom": {"sys": {"type": "Link", "linkType": "Upload", "id": "UPLOAD_ID"}}
        }
      }
    }
  }'

# Step 3: Process the asset
curl -s -X PUT "https://api.contentful.com/spaces/SPACE_ID/environments/master/assets/ASSET_ID/files/en-US/process" \
  -H "Authorization: Bearer CMA_TOKEN" \
  -H "X-Contentful-Version: VERSION"

# Wait 10-15 seconds for processing

# Step 4: Publish the asset
curl -s -X PUT "https://api.contentful.com/spaces/SPACE_ID/environments/master/assets/ASSET_ID/published" \
  -H "Authorization: Bearer CMA_TOKEN" \
  -H "X-Contentful-Version: NEW_VERSION"

# Step 5: Link asset to blog entry (update the media field)
# Fetch current entry, update media field, PUT, republish

9. Format 3: Social Media Thread

Every study produces a 3–5 post social thread. The thread drives traffic to the article and study share link.

Thread Structure Template

# Post 1: Hook (with infographic image attached)
We asked 10 [demographic] about [topic].

[Boldest finding as a statement].

Here's what they told us 👇

# Post 2: Key finding with data
[X] out of 10 said [finding].

The most common frustration? "[Direct persona quote]"

# Post 3: Surprise insight
But here's what surprised us:

[Unexpected finding or contradiction]

One participant put it perfectly: "[Vivid quote]"

# Post 4: Implication
What this means for [industry/role]:

• [Bullet 1: actionable implication]
• [Bullet 2: actionable implication]
• [Bullet 3: actionable implication]

# Post 5: CTA with links
📖 Full article: [article URL]
🔬 Explore the live study: [share link with UTM]

[2-3 relevant hashtags]

Thread Quality Rules

• Post 1 must hook. Lead with the most surprising or compelling finding. Attach the infographic image.
• Include at least 2 direct persona quotes. These are more engaging than summarised findings.
• Data points beat adjectives. "7 out of 10 said" beats "most people think."
• Keep each post under 280 characters (Twitter/X limit). Brevity forces clarity.
• Final post always includes both links: article URL and study share link.

Twitter/X API Posting

# Post tweet with image (Python example):
import requests

def post_tweet_with_image(text, image_path, api_credentials):
    """Post a tweet with an attached image."""
    # Step 1: Upload media
    media_upload_url = "https://upload.twitter.com/1.1/media/upload.json"
    with open(image_path, 'rb') as img:
        media_resp = requests.post(
            media_upload_url,
            auth=api_credentials,
            files={"media": img}
        )
    media_id = media_resp.json()["media_id_string"]

    # Step 2: Post tweet with media
    tweet_url = "https://api.twitter.com/2/tweets"
    tweet_resp = requests.post(
        tweet_url,
        auth=api_credentials,
        json={
            "text": text,
            "media": {"media_ids": [media_id]}
        }
    )
    return tweet_resp.json()

# Step 3: Reply to create thread
def reply_to_tweet(text, reply_to_id, api_credentials):
    """Reply to a tweet to continue the thread."""
    tweet_url = "https://api.twitter.com/2/tweets"
    return requests.post(
        tweet_url,
        auth=api_credentials,
        json={
            "text": text,
            "reply": {"in_reply_to_tweet_id": reply_to_id}
        }
    ).json()

10. Format 4: Email Outreach with Study Link

Email content uses study findings to create personalised, value-first outreach. The email delivers an insight before asking for anything. The study share link provides proof.

Email Structure Template

# Email structure (150-300 words):

Subject: [Specific finding] — we studied [topic] for [company/industry]

[Name],

[Opening hook: 1-2 sentences with the most relevant finding for THIS
recipient. Reference their company, product, or industry specifically.]

We ran a study with 10 [demographic description] and asked them about
[topic relevant to recipient]. Here's what stood out:

• [Finding 1 — most relevant to recipient's business]
• [Finding 2 — surprising or counterintuitive]
• [Finding 3 — actionable for recipient]

[One vivid persona quote that would resonate with recipient]

You can explore the full study here: [share_link with ?utm_source=ce]
(The link text should NOT show the UTM parameter)

[Soft CTA: "Would this kind of research be useful for [their upcoming
launch/campaign/strategy]? Happy to walk through the methodology."]

[Sign-off]

Personalisation Strategy

The same study produces different email angles for different recipients:

Email Sending via SendGrid

# Python: Send personalised outreach email
import requests

def send_outreach_email(to, subject, plain_text, html_content, categories):
    """Send email via SendGrid with tracking."""
    response = requests.post(
        "https://api.sendgrid.com/v3/mail/send",
        headers={
            "Authorization": "Bearer SENDGRID_API_KEY",
            "Content-Type": "application/json"
        },
        json={
            "personalizations": [{"to": [{"email": to}]}],
            "from": {"email": "[email protected]", "name": "Your Name"},
            "subject": subject,
            "content": [
                {"type": "text/plain", "value": plain_text},
                {"type": "text/html", "value": html_content}
            ],
            "tracking_settings": {
                "open_tracking": {"enable": True},
                "click_tracking": {"enable": True}
            },
            "categories": categories  # e.g., ["outreach", "sender-name", "cpg-sales"]
        }
    )
    return response

# IMPORTANT: Always include sender category in categories list
# IMPORTANT: Log the full email content in your database BEFORE sending
# SendGrid does NOT store email bodies — your database is the only record

11. Formats 5–7: Executive Summary, Sales Leave-Behind, Press Release

Format 5: Executive Summary (1 Page)

# Executive Summary Template:

## [Topic] — Research Summary
### [Date] | [X] Participants | Ditto Synthetic Research

**Key Finding:** [One sentence headline finding]

**Context:** [2-3 sentences on why this research was conducted]

**Top 3 Findings:**
1. [Finding with data point] — "[Supporting persona quote]"
2. [Finding with data point] — "[Supporting persona quote]"
3. [Finding with data point] — "[Supporting persona quote]"

**Implications:**
• [Actionable recommendation 1]
• [Actionable recommendation 2]
• [Actionable recommendation 3]

**Explore the full study:** [share link with ?utm_source=blog]

---
Methodology: Ditto synthetic research, [X] AI personas,
[demographic description], [X] questions. 92% correlation
with traditional research validated by EY.

Format 6: Sales Leave-Behind (Customer-Facing)

# Sales Leave-Behind Template:

## What [Target Audience] Really Think About [Category]
### Independent Research by [Your Company]

We asked [X] [demographic description] about [topic].
Here's what they said — in their own words.

**The #1 frustration:** "[Direct quote from Q2]"
(Cited by [X] out of [Y] participants)

**What they want:** "[Direct quote from Q3 or Q7]"

**What would make them switch:** "[Direct quote from Q6]"

**The surprise:** [Most unexpected finding from the study]

---
This research was conducted independently using Ditto synthetic
research methodology. Explore the full study and every individual
response: [share link with ?utm_source=ce]

Format 7: Press Release (Media-Ready)

# Press Release Template:

## [Company Name] Research Reveals [Headline Finding]
### New study of [X] consumers/professionals finds [key data point]

[CITY, DATE] — [Company Name] today released findings from a new
research study exploring [topic]. The study, conducted with [X]
[demographic description] using Ditto's synthetic research
methodology, found that [headline finding with data point].

**Key findings include:**

• [Finding 1 with data point]
• [Finding 2 with data point]
• [Finding 3 with data point]

"[Most quotable persona response]," said one study participant.

[1-2 paragraphs of context and implications]

**About the study:** [Methodology description]
**Full study results:** [share link with ?utm_source=blog]
**About [Company Name]:** [Boilerplate]
**Media contact:** [Contact details]

12. The Always-On Research Calendar

A content marketing engine runs continuously, not in bursts. The research calendar determines what studies to run and when.

Monthly Content Production Target

# Realistic monthly content engine output:
#
# 4 x monthly pulse checks   = 4 blog posts, 4 social threads
# 1 x quarterly deep dive    = 1 cornerstone + 6 derivative formats
# 1-2 x ad hoc probes        = 1-2 timely articles
#
# Total: 6-7 articles/month + social + email + sales content
#
# Time investment: ~8-12 hours/month
# Traditional equivalent: ~$20K-$40K/month in agency fees

Content Topic Selection

Use this framework to select study topics that maximise content value:

13. Worked Example: DTC Skincare Brand Content Engine

curl -s -X POST "https://app.askditto.io/v1/research-groups/recruit" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Content Study - Skincare Purchase Decisions - Feb 2026",
    "description": "Women aged 25-45 interested in skincare and beauty products",
    "group_size": 10,
    "filters": {
      "country": "USA",
      "age_min": 25,
      "age_max": 45,
      "gender": "female",
      "employment": "employed"
    },
    "sampling_method": "random",
    "deduplicate": true
  }'

14. Measuring Content Performance

Track these metrics to evaluate whether your content marketing engine is producing results. Maintain a 60/40 split between leading indicators (engagement) and lagging indicators (revenue impact).

Leading Indicators (60%)

Lagging Indicators (40%)

15. Best Practices and Common Mistakes

Do

• Run the study BEFORE writing the article. The study is the content. An article written first and then "supported" by a study link is generic content with a study appended. The study findings must drive the narrative.
• Use 10 personas minimum per study. Fewer produces findings that lack the pattern diversity needed for compelling content. "3 out of 10 said X" is weak. "8 out of 10 said X" is a headline.
• Ask questions sequentially. Conversational context produces richer, more quotable responses. Batching questions produces shallow, disconnected answers.
• Harvest language directly from persona responses. Use their exact words for headlines, email hooks, and social posts. Customer language resonates more than marketing copy.
• Generate all 7 formats from every deep-dive study. The marginal effort is small but the content multiplication is substantial. A study that only produces a blog post is underutilised.
• Include the study share link in every format. The share link is proof. Blog articles, emails, social posts, sales leave-behinds, and press releases should all link to the live study. Always include UTM parameters.
• Optimise for both SEO and GEO. Fill in all metadata fields: seoDescription, llmSummary, keyTakeaways, primaryClaim, quotableInsights. Articles without GEO metadata are invisible to AI search systems.
• Tier your content by study depth. Monthly pulses produce short articles (800–1,200 words). Quarterly deep dives produce cornerstone articles (1,500–2,500 words). Not every study needs all 7 formats.
• Save all raw responses. Quote banks have compounding value. The best social post from a study may emerge weeks later when you re-read the responses with fresh context.

Don't

• Don't publish articles under 1,000 words. Research-backed articles need enough depth to justify the "research-backed" claim. Thin content from rich data is a waste.
• Don't skip the infographic. Visual content dramatically increases sharing and time-on-page. The three-phase workflow (stock fallback → generate → replace) ensures the article is never blocked by infographic generation time.
• Don't use generic questions. "What do you think about our product?" produces generic content. "What's the ONE thing that would stop you from trying this?" produces specific, actionable insight.
• Don't publish without GEO metadata. The llmSummary and keyTakeaways fields are what make your content visible to AI search systems. Skipping them is leaving discoverability on the table.
• Don't recycle study findings across multiple articles. Each article should be based on a unique study. The entire point of the engine is that you can produce fresh primary research at scale. Recycling findings defeats the purpose and dilutes SEO value.
• Don't forget to call /complete on the study. Study completion triggers Ditto's analysis (segments, shared mindsets, divergences). This analysis provides additional content material and a richer share link experience.
• Don't batch questions. Every guide in this series repeats this because it is the single most common mistake. Sequential questioning produces dramatically better content.
• Don't track vanity metrics alone. Blog views without conversion data are meaningless. Study share link clicks without pipeline attribution miss the point. Always connect content metrics to business outcomes.

Common API Errors

16. Frequently Asked Questions

How long does the full workflow take from study design to 7 published formats?

45–90 minutes for the core workflow (study + blog article + infographic + social thread). The remaining formats (executive summary, sales leave-behind, press release, email) add 10–15 minutes. Total: approximately 60–105 minutes for all 7 formats from a single study.

How many articles can I realistically produce per month?

With 4 monthly pulse checks and 1 quarterly deep dive, expect 5–7 articles per month with approximately 8–12 hours of total effort. Teams that run additional ad hoc studies for trending topics can reach 10–15 articles per month.

What makes this content better for SEO than normal blog posts?

Three things: (1) original primary research data that no competitor has, (2) structured metadata (keyTakeaways, quotableInsights) that search engines parse, and (3) unique persona quotes that add E-E-A-T signals (experience, expertise, authoritativeness, trustworthiness). Google's algorithms explicitly reward original research over derivative content.

What is GEO and why does it matter?

GEO (Generative Engine Optimisation) is the practice of structuring content so AI systems (ChatGPT, Claude, Perplexity) can find, understand, and cite it. The llmSummary, keyTakeaways, primaryClaim, and quotableInsights fields make your content machine-readable and citation-ready. Without GEO optimisation, your content exists for human readers only and is invisible to AI-mediated discovery.

Can I reuse a research group across multiple studies?

Yes. If the same audience is relevant to multiple content topics, create a new study referencing the same research_group_uuid. Personas retain context from previous studies, which can produce deeper insights. However, for content that needs fresh perspectives, recruit new groups.

How should I choose between a 3-question pulse and a 7-question deep dive?

Use 3-question pulses for quick reads on shifting sentiment ("How do consumers feel about X this month?"). Use 7-question deep dives when you need comprehensive content material: the full narrative arc, competitive context, quotable moments, and actionable implications. Deep dives produce cornerstone articles; pulses produce timely updates.

Should I publish the article before the infographic is ready?

Yes. Follow the three-phase workflow: publish with a fallback image, generate the infographic, then replace and republish. This ensures the article goes live immediately while the infographic generation (2–5 minutes) runs in the background.

What if the study findings are not interesting?

This happens rarely when questions are well-designed. If findings are bland, look for: (1) minority opinions that contradict the majority — these make good "but here's the twist" moments; (2) unexpected language in persona quotes — vivid phrasing makes even moderate findings compelling; (3) the gap between what personas say they want and what they actually do — this contradiction is always interesting. If truly no compelling angle exists, the study still provided valuable learning; not every study needs to become content.

How does this compare to traditional content marketing?

Traditional: commission research from an agency ($5K–$15K), wait 4–6 weeks for results, hire a writer ($500–$2K), hire a designer ($500–$1K), spend 2 weeks in editorial review. Total: 6–10 weeks, $6K–$18K, one article. Ditto + Claude Code: run study (20 min), generate 7 formats (70 min). Total: 90 minutes, API cost only, seven formats.

Can I use this for B2B content marketing?

Yes. B2B content benefits even more from original research than B2C, because B2B audiences are more sceptical and data-driven. Adjust the group description to specify professional context ("Senior product managers in mid-market SaaS companies"), use education and employment filters to target professionals, and customise questions for B2B problem spaces.

Related guides:

• Positioning Validation Guide for Claude Code
• Competitive Intelligence Guide for Claude Code
• Product Messaging Guide for Claude Code
• Pricing Research Guide for Claude Code
• Customer Segmentation Guide for Claude Code
• Voice of Customer Guide for Claude Code
• Sales Enablement Guide for Claude Code
• Ditto API for Claude Code: Complete Guide
• Installing the Ditto Skill

Ditto — Synthetic market research with 300,000+ AI personas. Validated by EY (92% correlation), Harvard, Cambridge, Stanford, and Oxford.
askditto.io · [email protected]