June 1, 2023

ChatGPT for Marketers — A Complete Guide

Reading time about 16 min

ChatGPT is a game changer for marketers short on time. This guide goes deep into ChatGPT meta to show you how to construct prompts that generate quality, original content for any marketing use case.

ChatGPT updates: What’s new

ChatGPT 4 is out now. It’s available in beta for ChatGPT Plus subscribers. ChatGPT 4 will soon be multimodal, which means it’ll be able to generate images and videos as well as text output. 

Plugins are now supported. Many third-party apps have already introduced end-to-end task automation through AI agents, which can browse the internet on their own, leaving marketers free to focus on other tasks. 

Read on to learn how you as a marketing expert can use all these new features and ChatGPT-powered tools to boost your productivity.

What is ChatGPT?

ChatGPT is a large language model (LLM) trained by OpenAI. It uses GPT (Generative Pre-trained Transformer) architecture, a type of artificial intelligence that identifies language patterns and uses those patterns to figure out how to generate new text.

The “pre-trained” part of an LLM means it’s trained on a large amount of data before it’s used for a specific task. ChatGPT in particular is trained on a big chunk of the internet and other publicly sourced data, including Wikipedia, books, and videos (via transcription). That’s why it can answer questions about almost anything.

ChatGPT interface prompting user to type something.

source

ChatGPT 4 is the latest conversational AI model from OpenAI. The cutoff date for the current dataset ChatGPT is trained on (both versions 3.5 and 4.0) is September 2021. 

Anything published after the cutoff date (including news, research and important events) would be outside of ChatGPT’s current scope of knowledge. 

The default ChatGPT interface can provide search-enriched results through Bing and other internet-connected ChatGPT plugins. However, those features are only available to ChatGPT Plus subscribers at the moment.

What is the difference between Microsoft Copilot and ChatGPT?

Microsoft Copilot (formerly Bing Chat) is an AI search tool provided by Microsoft. It uses a chat-style interface and can be accessed through the Edge Browser, Bing for Android, or Bing for iOS

Microsoft Copilot is free (with ads) and runs a customized version of ChatGPT 4. It can generate answers to queries based on internet search results, write long streams of text, and even code.

Microsoft Copilot has filters and adjustments in place that occasionally make its output noticeably lower quality from ChatGPT, albeit much more accurate most of the time. 

Microsoft Copilot has some restrictions:

  • 4000 characters per query limit
  • 20 questions limit per conversation thread
  • 200 queries per day quota

In contrast, the default ChatGPT interface provides unlimited conversations, no restrictions on the length of each conversation thread, and a slightly higher input limit of 4096 characters per query.

ChatGPT as provided by OpenAI lets you choose between different versions of its model. ChatGPT 3.5 is available for free for all users, and ChatGPT 4 is only available to ChatGPT Plus subscribers (at $20/per month). 

ChatGPT Plus subscribers can also perform internet searches with results drawn from the Bing search engine. As of writing this article, the complementary ChatGPT internet search function is much slower, less accurate, and less powerful than Microsoft Copilot (formerly Bing Chat).

Microsoft Copilot modes and the Bing sidebar

Microsoft Copilot lets you choose between two distinct modes: Chat and Compose. 

Chat mode is on by default and takes a question-and-answer approach, using the internet to search for correct answers. It’s conversational but can also accept short-form writing tasks. 

Example of Bing Chat writing marketing content.

Compose mode is designed for content generation and is only available in the sidebar of the Edge Browser. 

You can access it by clicking on the Bing logo of the Edge browser.

Bing chat compose sidebar with settings.

Compose mode in Microsoft Copilot lets you adjust the tone, format, and length of the content you want to generate, which makes it especially useful for copywriters. 

You can ask Microsoft Copilot to answer questions or generate content based on the webpage you currently have open in your browser.

This makes Microsoft Copilot a goldmine for marketers who want to delve deeper into competitive analysis, save time on market research, or produce short-form content for blog and social media posts.

Microsoft Copilot or ChatGPT?

TLDR: If you want to speed up your research and get up-to-date info, Copilot is your best bet. If you want to work with a lot of text and have a back and forth longer than 20 responses (Copilot’s current limit), then use ChatGPT’s default online interface.

To decide when to use Microsoft Copilot or ChatGPT, consider their different strengths and weaknesses.

Copilot is better at research, built to answer questions as you browse the internet, with useful features like context awareness and automatic source citation. 

Microsoft Copilot is weak in any context that requires originality or creativity, and can’t handle complex instructions.

ChatGPT is better at content generation. With a well-constructed prompt, it can produce high quality, original content.

Bear in mind that while Microsoft Copilot runs on ChatGPT 4, the results it generates can be vastly different from the default ChatGPT app because of the limitations and guardrails that Microsoft has put in place. 

The benefits of AI marketing

AI marketing is any use of AI in the context of marketing. It could be a way to draw insights and gather suggestions from existing content like a web page or PDF document. It can also create something from scratch like a blog article or variations of text such as email subject lines.

Using ChatGPT for marketing and other AI tools is a no-brainer. AI is to marketers what a jetpack is to rock climbers. No matter what type of use case: ideation, research, comparison, insights, or content generation, you will save a ton of time and effort.

With the upcoming multi-modal capabilities, ChatGPT will eventually be able to parse and understand images and videos, including the text it sees in them — like turning a sketch on a napkin into working code for a website.

The drawbacks of AI marketing

The most important AI-related legislation to date, the EU Artificial Intelligence Act, was drafted before large language models like ChatGPT and competitors Claude, Open Assistant, and LaMDA (which powers Google Bard) were widely used, making most of the governing rules on the use of AI irrelevant to marketers, and of little protection to regular users in general.

What matters today for AI marketing is legal compliance with existing laws like not sharing user data with third parties without express permission, and not using AI marketing apps or AI tools to commit plagiarism, fraud, or other crimes. 

Legal and ethical concerns when using ChatGPT

OpenAI is headquartered in San Francisco, California and therefore strictly complies with US regulations, making it safe to use their tools for commercial work. Third-party apps powered by the ChatGPT API will have their own separate terms of use that require individual consideration.

With the above in mind, the biggest issue that still plagues Open AI’s most popular AI models (ChatGPT, DALL·E, and Whisper) is how much of the openly accessible content on the internet was used as training data.

To what extent prior datasets used to pre-train ChatGPT 3.5 and 4 were copyrighted or now retroactively disallow use for LLM training is unknown, which raises ethical and legal questions.

OpenAI (and other leading AI companies) have since made efforts to legally obtain permission to include copyrighted content in newer datasets. However, user submitted input, which applies to any and all information you feed ChatGPT is still actively used to train OpenAI’s LLMs, which raises serious privacy concerns.

Some of the privacy concerns leveled at OpenAI have been addressed by its new ChatGPT user data management system, which lets you disable chat history and prevent OpenAI from using your user-submitted data to further train versions of its ChatGPT models.

This doesn’t mean you can provide ChatGPT with sensitive information, such as company records or legally-protected user data. Even with chat history turned off, you still run the risk of breaking major EU and US privacy laws. 

The EU General Data Protection Regular (GDPR) for example already protects users against unwarranted sharing of personal information with third parties, and that extends to third-party tools like ChatGPT. 

For now, you can turn off chat history when providing sensitive company information, and use dummy content or anonymized user data to ensure you don’t run afoul of local and international data privacy laws.

How does ChatGPT work? And how smart is it?

As many machine-learning experts have pointed out, ChatGPT is a predictive AI model that analyzes patterns and tries to guess what word comes next when formulating a response.

In theory, ChatGPT doesn’t really understand what it’s doing, which often leads to what some experts call hallucinations. 

AI hallucinations occur when ChatGPT makes stuff up, going as far as citing false information with utmost confidence.

Because of this phenomenon, the consensus among AI experts is that LLMs do not currently possess the logic and reasoning skills of a human being.

Despite lacking human-level awareness, the current performance of modern AI is so outstanding, it already outperforms human experts in highly technical fields.

For example, ChatGPT 4 has shown significant improvements in reasoning and logic, matching top performers on specialized tests such as bar exams and science olympiads. 

The mystery behind why AI suddenly became much more powerful is often attributed to a concept called ‘emergence’, where new, unforeseen capabilities emerge as an AI model is further developed. 

All large language models have emergent qualities, including ChatGPT. This means that we don’t yet know the limits of LLMs or AI tools in general, as AI technology is still in its infancy.

The right way to use ChatGPT as a marketer

Quality input is key to the performance of AI tools like ChatGPT. Prompts, as they’re called, are expertly crafted questions, designed to garner the best answers from ChatGPT and other popular generative AI models like Midjourney.

ChatGPT marketing use cases

ChatGPT can be broadly used in any AI in digital marketing context. The most popular use cases for ChatGPT marketing include:

  • Writing marketing copy (from subject lines to blog articles and more)
  • Generating ideas for marketing campaigns
  • Creating content outlines and comprehensive guides
  • Analyzing patterns to offer insights and predict trends
  • Summarizing long-form content (such as a PDF or a webpage)
  • Rewording content to fit a particular tone of voice or style
  • Repurposing content (like creating social media posts from a podcast transcript)
  • Fixing grammatical issues and improving readability
  • Filtering and organizing content into new formats (like a table or csv list)
  • Optimizing marketing workflows through automations

How to create effective ChatGPT marketing prompts

To get the best results when using ChatGPT for marketing, your marketing prompts need enough information and a well-explained context.

Here are five steps to take to create effective prompts when using ChatGPT for marketing:

1. Provide sufficient context

In prompt design, context includes the “who”, “what”, and “why”.

Start with roleplay, your requirements, and your expected outcomes to provide ChatGPT with a complete context. 

Step 1 — Roleplay: “As an SEO expert with experience in running effective ecommerce campaigns for globally recognized brands…” 

Step 2 — Requirements: “Draft a 1,000-word article on the subject of Mother’s Day gifts with the following keywords…”

Step 3 — Outcomes: “This article should encourage last-minute shoppers to buy a Mother’s Day gift to impress the mom in their family…”

2. Define a mental model

A mental model is a way of doing things. It’s the “how” in prompt design.

Include specific rules, conditions and techniques for ChatGPT to follow. Define the tone of voice, writing style, audience perspective, and marketing technique. 

Step 1 — Rules: “Avoid overly salesly language and promotional exaggerations. Exclude the following words and phrases…” 

Step 2 — Conditions: “Write exclusively for native English speakers with a friendly tone of voice and informal writing style…”

Step 3 — Technique: “Using the AIDA marketing model, assume our target audience is at the interest stage…”

3. Include relevant data

Take advantage of whatever information you have available.

Share anonymized user data or market research, alongside examples of similar marketing content and specific product details.

Step 1 — Demographics: “Our audience demographics are married couples with children, aged 20 to 39, with 70% residing in North America…” 

Step 2 — Competitors: “Our main competitors include the following brands and their key selling points…”

Step 3 — Products: “Our products include the following, along with a description of our brand and its key selling points…”

4. Give structured feedback

Structure the process of receiving and sharing feedback before you start.

Include stakeholder input and any prior feedback. Let ChatGPT create its own list of questions and requirements to improve content generation. 

Step 1 — Stakeholders: “Keep in mind the following requirements from the product and branding teams…” 

Step 2 — Feedback: “Here is some content I want you to include with slight tweaks based on the guidelines above…”

Step 3 — Questions: “Before you start on the above task, create an outline for your process and ask any questions that will help produce better content.”

5. Keep refining your results

Ask ChatGPT to change specific phrases, sentences or paragraphs. 

Specify whether you want a rewrite or new variations. When asking for rewrites, ask ChatGPT to list the changes in each section.

Step 1 — Changes: “Here is a list of changes to make to the above. Leave everything else unchanged…” 

Step 2 — Options: “Create 3 variations of the following sections while strictly adhering to the above instructions…”

Step 3 — Questions: “Whenever you make a change, indicate where it was made after each section…”

ChatGPT-powered apps for marketers

Building your own AI workflow for marketing can be time-consuming. Luckily, there are plenty of apps that extend the functionality of ChatGPT to further improve your content creation process.

The best ChatGPT powered apps for content marketing include ChatSonic and Jasper Chat which both provide fine-tuned workflows for specific marketing tasks. 

If you prefer to use free tools, adding a ChatGPT browser extension in Edge, Chrome or Opera (which all support chrome extensions) lets you use ChatGPT with an internet-connected experience.

Browser extensions like ChatGPT Sidebar allow users to generate content based on an internet search or what they see on any open web page, similar to Microsoft Copilot sidebar experience in the Edge browser but without the limitations imposed by Microsoft. 

For help with prompt ideas from a community of over 900K users, AIPRM for ChatGPT is by far the best ChatGPT browser extension for power users, and can provide help with any kind of marketing task involving content generation.

When you want to analyze long-form content that exceeds the input limit in ChatGPT, use apps like ChatPDF, MapDeduce and ChatDoc that let you upload a document and ask questions based on its contents.

For video content, you can use summarize.tech to summarize any YouTube video. Just upload your own video as ‘unlisted’ on YouTube to use this tool on your content while keeping your video private. 

For in-depth research, especially work that needs to cite scientific and academic papers, Genei and Elicit are must-haves.

Lastly, consider using AI agents like AutoGPT as fully automated task managers that can regularly perform a series of steps based on prior instructions.

AI agents are a great way for marketers to automate more complex tasks, such as automatic blogging, keyword research, and monitoring the web for updates on specific topics.

You can quickly set up and deploy your own AI agent without any technical know-how using AgentGPT, AI Agent or Aomni.

How to automate your marketing efforts

Marketing automations help you do more in less time. Ready to get started?

Brevo’s AI features empower you to work efficiently and create impactful campaigns. Use the AI assistant to generate subject lines, email content, and more.

Ready to grow with Brevo?

Get the tools you need to reach your customers and grow your business.

Sign up free