Turn Dorqa data into personas, segments and post ideas in Claude and ChatGPT

Four prompts that turn Dorqa data into a buyer persona, audience segments and post ideas in Claude and ChatGPT, each written to stop the model inventing things.

The short answer

You can ask Claude or ChatGPT for a buyer persona, audience segments or post ideas, and get answers built on real data.

Connect Dorqa first. Then paste one of the four prompts below.

Why it works: Dorqa gives the assistant the questions your audience really asks. Every claim comes with a link you can open.

Claude can run all four prompts. ChatGPT can run one, post ideas from a trend.

What can you make with it?

Prompt 1: a buyer persona. Who you are writing for.

Prompt 2: audience segments. The different types of buyer inside one audience.

Prompt 3: post ideas from a trend. Seven posts you can write this week.

Prompt 4: trend plus audience. Posts your audience wants, on news moving right now.

Claude or ChatGPT?

Claude: all four prompts.

ChatGPT: prompt 3 only, for now.

Why the difference? Prompts 1, 2 and 4 open an audience's research. That uses Dorqa credits, and paid tools are not in ChatGPT yet. They are planned.

How do you connect?

You need a Dorqa account first. Sign up free.

For Claude:

Step 1. Open Settings, then Connectors, then Add custom connector.

Step 2. Name it Dorqa. Paste https://mcp.dorqa.io/mcp

Step 3. Press Connect and sign in to Dorqa.

Step 4. Tick "Allow this assistant to spend credits". It is off by default. You need it for prompts 1, 2 and 4.

For ChatGPT:

Step 1. Search the apps directory for Dorqa and connect it.

Step 2. Sign in with your Dorqa account. ChatGPT reads free data only.

To change the spending choice later, disconnect Dorqa and connect again.

What does it cost?

Trends, questions and the list of researched audiences are free, in both assistants.

Opening an audience for the first time uses credits. Opening it again is free, here and in the Dorqa app.

The Free plan includes one audience, for life.

To see the price, ask the assistant: "What plan am I on and what does opening an audience cost?" It reads your live balance and prices.

There is also one daily credit limit for assistants and API keys together. A runaway chat cannot use up your balance.

Where do audiences come from?

The audience library is shared by every Dorqa user. If yours is not there yet, research it once in the Dorqa app. After that, Claude can open it.

Want to see what an audience report looks like? Read What the Search Data Reveals About Digital Marketers in India. Curious how the research works? Read How We Built Audience Intelligence.

Six rules every prompt follows

Each prompt below already has these built in. Keep them if you write your own.

1. Cite or skip. Every claim links to its source. No link, no claim.

2. Obey the labels. Some data is measured in your market. Some is global. Never call global data local.

3. No skew without a baseline. Gender data exists for the US and UK only. Elsewhere, say nothing about gender.

4. Drop the noise. Skip URL fragments, community names listed as pain points and off-topic keywords.

5. Say what is missing. An empty field is "not in the data". Never a guess.

6. Date it. Every answer states the snapshot date.

Prompt 1: Buyer persona

Use it when you need to know who you are writing for. Works in Claude. Replace the words in square brackets.

Use Dorqa. Find "[audience]" with list_audiences. Before you open it with get_audience, tell me what it will cost and wait for my OK. Then run get_audience for the [market] one.
Build one buyer persona with these parts: who they are, what they want, what worries them, where they gather, who they trust, and the questions they ask.
Rules:
Use only what the data contains. Every claim cites the question, thread or community behind it, with its link.
Obey each section's market_scope and how_to_read notes. Never call a global section local.
Never state a gender skew where gender_over_index is missing.
Skip anything that looks like noise: community names listed as pain points, URL fragments, off-topic keywords.
If a part has no evidence, write "not in the data".
End with the snapshot date and three things this data cannot tell me.

What you get: for digital marketers in India, the persona leads with money and career doubt. Real questions in the data: "Can I earn 10 lakh per month from digital marketing?" and "Do digital marketers get paid well?"

Prompt 2: Audience segments

Use it right after the persona, in the same chat. One audience is rarely one buyer. Works in Claude. It costs nothing extra, because reopening the same audience is free.

From the same get_audience data, split this audience into 2 to 4 segments.
Base the split only on differences you can see in their questions, threads and communities. For example: people asking how to start, versus people asking about pay or clients.
For each segment give: a short name, the evidence (at least two items, each linked), where they gather, and one offer that fits them.
No segment without evidence. Same rules as before: cite or skip, obey market_scope, drop the noise.

Tip: if a segment has only one weak piece of evidence, ask the assistant to merge or remove it. Fewer, stronger segments are more useful.

Prompt 3: Post ideas from a trend

Use it when you need something timely to post this week. Works in Claude and ChatGPT. It uses no credits.

Use Dorqa. Run list_trend_categories and pick [category]. Run search_trends in that category. Pick the trend with the most signals_30d that is not marked stale. Run get_trend on it.
Give me 7 post ideas. Each one is built on one of its recent_signals articles, with the link: 2 LinkedIn text posts (hook plus 3 key points), 2 carousel outlines (slide titles), 2 short video hooks (first line only), and 1 contrarian take.
Do not quote stage or velocity as fact. Say "covered by N publishers in the last 30 days" instead. Do not copy sentences from the articles. Write in your own words and link the source.

Why the link matters: a post that says "Ahrefs tracked 37,500 AI citations and found..." with the source link is worth more than a generic take. Your reader can check it. So can AI search.

Prompt 4: Match a trend to your audience

The most useful of the four. It finds posts your audience wants, on news that is moving right now. Works in Claude.

Use Dorqa. Run get_audience for "[audience]" in [market], and get_trend for "[trend]". If opening the audience costs credits, tell me the price first and wait for my OK.
Match the trend's recent_signals articles to the questions and threads this audience has. For each match give: the audience question (linked), the article that answers it (linked), and one post hook that answers the question using the article, in my own words.
Only real matches. If an article answers no question in the data, leave it out.

A worked example

Audience: digital marketers, India. Trend: AI Search Visibility. Data as of 30 October 2026.

Match 1.

Audience question: "Is digital marketing a dead field?" From a Reddit thread in India.

Article that answers it: "AI-cited pages more often got straight to the answer", Search Engine Journal.

Post hook: Digital marketing isn't dead. How you get found changed. AI now cites the pages that answer first.

Match 2.

Audience question: "Do digital marketers get paid well?" From Google People Also Ask.

Article that answers it: "How stable are AI citations?", Ahrefs.

Post hook: The marketers who earn more next year can explain why AI cites one page and not another. Most can't yet.

What can't it do?

It cannot open audience research from ChatGPT yet. Use Claude for prompts 1, 2 and 4.

It cannot research a brand-new audience from either assistant. Research it once in the Dorqa app. Then Claude can open it.

It cannot measure everything in your market. Search volume, questions and age interest are measured in your market. Reddit, YouTube and the language people use are global content, ranked from your market.

It cannot hear audiences that search in a local language. All searches run in English, so those audiences are under-counted.

It cannot replace talking to customers. A persona from public data tells you where to start. Ten customer calls tell you if it is right.

It cannot write the post for you. The hooks are a first draft. Your voice is what makes it worth reading.

Try it now

Start free on Dorqa, connect Claude or ChatGPT, and run prompt 3 first. It is free and works in both. See plans and pricing when you want audience research in Claude.

Be first. Not just found.

Hemant Sharma
Hemant Sharma
Founder, Dorqa
Hemant Sharma is the founder of Dorqa, an SEO and AI Search platform. He has worked in SEO for nearly a decade and runs AdSchoolMaster, a community of 6,000+ marketers, and the SEO agency BirdMatrix.

Get the build log

One email when we publish — field notes on search, trends, and showing up in AI answers. No gated PDFs.

No spam. Unsubscribe anytime.

Be first. Not just found.

Join 7,000+ marketers using Dorqa for SEO, Trend Intelligence, and AI Visibility.

Start free — no card needed