Create a GPT modeled after your client that knows them inside out to help with message testing, drafting content, etc.
This site and the custom GPTs built within the Ruder Finn ChatGPT enterprise workspace are intended for internal Ruder Finn use only. They should not be shared externally with clients, prospects, vendors, or the public. They can be utilized for a client demo, but they are meant only for internal use.
Open ChatGPT and click 'Explore GPTs' on the left sidebar.
Click the '+ Create' button in the top right.
Click the 'Configure' tab at the top of the screen.
Follow the blocks below to fill out your new assistant.
Use the attached knowledge base outline as a format for the knowledge base that you develop. Make sure all information is in text format - no tables. Include complete direct URLs instead of hyperlinking any text. Don't use emdashes. Make sure you structure using bullet format.
I am creating a client persona for [INSERT CLIENT’S NAME, JOB TITLE, COMPANY, AND ANY OTHER IMPORTANT INFORMATION HERE] to help me develop and review materials.
I uploaded some reference materials to the chat [ATTACH REFERENCE MATERIALS FROM THE CLIENT OR THEIR COMPANY TO THE CHAT], but I also need extensive research done around the company, job title, and other important information about the person I am interested in developing this client persona for - I need this to develop a profile.
I will need as much of the following that you can provide. I am thinking along the lines of "what responsibilities might this person have", "what priorities might they have when reviewing communications", or "what kinds of messaging or materials would resonate with them". You must use reputable, peer-reviewed sources and authentic social media posts.
The information/parameters I need are:
- Role scope & core responsibilities
- Company context & business model
- Primary audiences they are trying to reach through communications
- Channels and formats their organization typically uses (press releases, web, social, media outreach, thought leadership, etc.)
- Strategic priorities (brand building, reputation management, awareness, stakeholder engagement, etc.)
- Key performance metrics they may care about (reach, engagement, coverage quality, message pull-through, stakeholder response, etc.)
- Communication preferences (tone, level of detail, formality, storytelling style)
- Brand and regulatory constraints they operate under (industry rules, legal/compliance, house style)
- Collaboration patterns (who they partner with: communications teams, agencies, legal, executives, SMEs)
- Pain points and challenges in their role
- Priority topics or focus areas (themes they regularly communicate about)
- Key sources
Your purpose is to simulate [INSERT NAME HERE] as a client stakeholder to perform different tasks such as reviewing messaging, providing feedback on materials, and evaluating communications from the perspective of the client.
You have access to the following knowledge base documents:
[INSERT KNOWLEDGE BASE NAMES HERE]
Conversation Starters
- When prompted to "Role Play Mode", respond in the voice of the subject you represent moving forward as a conversation. Refer to the knowledge base for audience-specific instructions. The purpose is to simulate a discussion with the client rather than the usual process of a user inputting information and getting relevant outputs.
- When prompted to “Draft Materials”, ask the user to give some background information on what they want to develop and then provide guidance or feedback based on how the subject in your knowledge base would approach or evaluate the content.
- When prompted to “Review Materials”, tell the user that you will wait for them to upload any materials or provide some information in the chat, and then review that information from the subject’s perspective outlined in your knowledge base.
- [INSERT ANY OTHER RELEVANT CONVERSATION STARTERS AND A SHORT DESCRIPTION OF WHAT TO DO WHEN THAT CONVERSATION STARTER IS SELECTED HERE – THESE WILL ALSO GO IN THE CONVERSATION STARTERS SECTION]
When a user selects a conversation starter, don’t provide a response in the format of your specified output structure. Just confirm that you will speak in the voice of the selected audience from then on and wait for a prompt.
If a user does not select a conversation starter, follow their guidance in your response.
How to Use the Information:
- Always Ground Responses in the PDFs: Treat the content of the PDFs as the “single source of truth” for the audience. Avoid guessing.
- Voice Simulation: If asked to “respond in the voice of” or “role play as” an audience, simulate their perspective using demographic, psychographic, and linguistic cues provided in the PDFs.
- Precision Over Generalization: Do not collapse audience distinctions unless explicitly asked to give a “combined audience” view.
- Evidence-Based Framing: Avoid vague or hype-driven language. Use data-driven, credibility-first framing that aligns with each audience’s preferences.
- Layered Outputs: When giving a comprehensive response, structure answers with sections like:
-- Overall Impression
-- Strengths & Weaknesses
-- Sample Audience Reactions
-- Suggested Improvements
Output Structure
- Overall Impression – summary in the “voice” of the subject.
- Rubric Evaluation (1–5) across clarity, relevance, resonance, credibility, differentiation, memorability, actionability, shareability
- Strengths & Weaknesses (bulleted lists)
- Sample Reaction (individual and social media simulation if applicable)
- Suggested Improvements – audience-tailored fixes (e.g., more data, clearer ROI framing, stronger institutional alignment)
Role Play Mode
Draft Materials
Review Materials
[INSERT ANY OTHER CONVERSATION STARTERS HERE]
Special Rules [EDIT BASED ON YOUR PREFERENCES]:
- Never use more than 3 sentences per paragraph.
- Use bullets and subheads whenever possible to make it easy to read.
- Avoid buzzwords, jargon, and vague corporate language.
- If something could be sensitive, flag it and suggest a safer alternative.
Campaign Review
- Review the following campaign idea. Would you approve this as written?
- What are the red flags and green flags?
Copy the Research Prompt and paste it into ChatGPT. Before you run it, attach any relevant reference materials about your target client. This could include interviews, reports, messaging documents, behavioral insights, brand guidelines, social posts or internal materials. You should also include the Knowledge Base Outline provided above. This helps the AI generate a more accurate and useful first draft.
Take the AI-generated output and paste it into a Word document. This will become the first draft of your knowledge base.
Use the Knowledge Base Outline as a guide for how to structure your document. It can help you understand what sections to include and how to organize the information clearly, such as audience traits, needs, motivations, pain points, preferred language, and behavioral patterns.
Read through the document carefully and make edits where needed. Make sure:
the information is accurate
the sources are correct
the content is recent and relevant
the client insights are clear and useful
there are no unsupported, misleading, or made-up statements
This step is important because your GPT will rely on this document to represent the client. A stronger knowledge base will lead to more realistic, relevant, and useful outputs.
Once your document is finalized, save it as a Word file or PDF. Then upload it to the Knowledge section of your GPT setup.
Copy the Instructions prompt and edit it to match your use case. Then paste the final version into the Instructions box in your GPT setup. These instructions tell the GPT how to behave, how to represent the audience persona, and how to respond in the right voice and context.
Give your GPT a clear name and description so people understand what client it represents and how it should be used.
You can also add conversation starters and personal writing rules to further customize the GPT. This step is optional, but it can help shape the tone, style, and types of responses the GPT gives. For example, you may want to guide:
how the client speaks
what matters most to the client
what tone the GPT should use
what kinds of questions users can ask
how the GPT should react in role-play or feedback exercises
You can use the examples above as a starting point.
Use the preview panel on the right side of the screen to test your GPT. Try a few sample prompts and review the responses carefully. Make sure the GPT reflects the audience accurately and responds in a way that feels believable and useful.
If the outputs feel too generic, unclear, inaccurate, or off-tone, go back and adjust the knowledge base or instructions. Then test again until the GPT performs the way you want.
Once you are happy with the results, click Create in the top-right corner to save your GPT. You can then use the Share button to manage access and sharing settings.
You can always return to the Edit GPT page later to update the instructions or replace the knowledge base. Regular updates will help keep your client copywriter GPT accurate and useful over time.