Marketing is based on empathy.
And empathy comes from research.
The best research comes from interviews.
But… stakeholder interviews take a lot of work. They’re hard to schedule and time intensive. It’s tough to connect with the right people. It takes skill to get them to open up. Often, you have just one chance.
Wouldn’t it be nice if you had unlimited time to talk? If they were available at a moment’s notice? If you could ask super direct questions, even the questions that make most people squirm?
That’s what an AI marketing persona is for.
You can use AI to create an artificial version of a member of your target audience and ask them anything about sales and marketing, triggers and topics, hopes and fears. And it’s always there for you, day or night, to guide your content and messaging.
Today we’re showing how to make your own AI marketing persona using Chat GPT and Bing. And to confirm that using AI to create personas isn’t a totally crazy idea, along the way we’ll check in with Ardath Albee, the leading persona expert.
A persona is a composite sketch of a target audience or audience segment describing your ideal customer. It guides your marketing strategy by accurately identifying the general traits, concerns and behavior patterns of your potential buyer. It’s a guide.
A good persona answers these types of questions:
You can create one simply by asking Chat GPT to create a first draft.
For our examples here, we’ll pretend we’re a social media marketing agency focused on helping nonprofits with fundraising. Our buyer is in a marketing leadership role and has fundraising and event promotion goals.
We’ll start with a detailed prompt. The best prompts help the AI by giving a lot of information. Our prompt will include the following elements of a persona:
We’ll also add to the prompt the specific output that we’re looking for in the response:
Here it is with the draft AI marketing persona. We’re including the prompts in their original form. Some of the responses have been edited for brevity. You’ll see my prompts and ChatGPT’s output.
Build a persona for a Marketing Director of a $10M non-profit in the midwest US with a mission to provide better access to healthcare resources in under-served communities. The goals of this person are to create more effective fundraising campaigns through digital marketing and also to promote their events. Include specific details about goals, pain points and decision criteria for selecting a social media marketing company.
Name: Taylor Rodriguez
Background: Taylor Rodriguez is a compassionate and dedicated Marketing Director of a non-profit organization, Midwest Health Initiative, which holds a mission to enhance access to healthcare resources in under-served communities. Their passion for public health started in college, where they majored in Public Health and minored in Marketing. Taylor later obtained an MBA to gain a stronger business foundation to effectively manage marketing campaigns for non-profits. Taylor has been working in the non-profit sector for 15 years, holding different roles ranging from outreach coordinator to marketing manager. Their lifelong commitment to healthcare access stems from personal experiences growing up in a marginalized community.
Decision criteria for selecting a social media marketing company:
That entire process took about two minutes using Chat GPT-4. It looks like a good start. But let’s show it to our friend Ardath Albee, a persona and audience research expert with 30+ years of experience in marketing strategy, and she was she thinks.
Ardath Albee, Marketing Interactions
“Do we really know this is accurate? If you incorporate this perspective into your content, what’s the likelihood that this is correct? Do these apply to your market? Your competitors’ market? You didn’t validate this. Check these with your customers before using this. I don’t trust this.“
She is understandably skeptical. And she’s right. We don’t really know if this is accurate. So as we proceed, we aren’t going to set our expectations high. We’ll plan to validate this through interviews with our actual customers.
ProTip: Never expect to get a great result from the first prompt.
You wouldn’t expect human intelligence to do great work after one simple request. The AI is no different. Expect to keep working with the AI, training it and improving the response with subsequent prompts.
Ardath recommends another marketing tool, Crystal, which you can use to evaluate the personalities of actual customers based on their online presence, and then look for patterns. In her experience, people with a clearly defined customer persona tend to have similar personality types.
As we proceed, we’ll continue to build up the target persona by adding details. We can do this by simply adding information that you have on hand, by using other tools or by continuing to prompt the AI.
Using the language (as in jargon) of the target audience creates alignment both for SEO and conversion. Marketers must always seek to use the words their audiences use.
I’ll show the prompt and response for that insight here:
What words does Taylor use to describe his marketing challenges and solutions?
Taylor, with their background in both public health and business, uses a unique mix of health sector-specific and general marketing terminology to articulate their challenges and solutions.
Those are some fancy words! Probably, I would have avoided that level of jargon, but if I can confirm that these are in fact, the words that my audience uses, I’ve struck gold. I’ll use them throughout our website, our campaigns and even our proposals.
Here, Chat GPT has major limitations. Today, as I write this, the native version at chat.openai.com doesn’t have the ability to browse the web.
But Bing does.
The Chat feature in Bing currently uses GPT-4 with web browsing built in. So it’s a better place to ask a question about sources of information. Here’s what the prompt and response look like in Bing:
I asked Bing where a person with this job title and company type would go to get information to help them do their job. It suggested several resources including events and degree programs. Useful subsequent prompts might include:
All of these are questions that can only be answered by an AI that can browse the current internet. And each of the responses can guide me toward the resources and topics that are relevant to my persona.
Each of these was insightful but not magic. Honestly, an audience intelligence tool such as SparkToro can give you even more insights even faster.
There’s a true story in the life of every lead. If you know where that story begins, you have a better shot at writing the ending.
Ask your AI buyer persona to tell you what was happening that made it realize that it needed your services.
What kinds of things might happen that make you realize that you need help from a marketing agency?
Several signals would alert me to the fact that our non-profit needs assistance from a marketing agency:
Because these are the initial drivers, then hitting these specifically will help win the business. All of the marketing materials (plus the pitch and proposal) should all tie back to the original ‘why.’
Again, let’s show this to Ardath for some expert insights from a human.
Ardath Albee, Marketing Interactions
“This is too high level. AI is great at making lists of stuff, but the problem is specificity. If you’ve actually done the work and have personas, then you can ask it to expand on them and you may get something of value. The question is what do you do with it?”
With another prompt or two, we can get ideas for what basic information we need to include in our marketing materials and sales conversations. Anything stories or examples that align with these should improve both our conversion rates and closing rates.
What information about the possible marketing agency would you need before contacting that marketing agency about a potential project?
Before contacting a marketing agency about a potential project, I would need to gather specific information to ensure the agency is a good fit for our needs. Here are the key points I would consider:
Although most of that information may seem general, it’s at least a checklist for topics and formats. It suggests that all messages should be reinforced with highly-relevant examples throughout the marketing and sales processes. And alignment on values, although not the most important thing, needs to be there.
The answers to these questions can help align your offers with their needs. What do they need and how do they buy it? Who is involved in the process? Repeated prompting about the buying process and the decision criteria can guide the sales strategy.
Which marketing functions are you most likely to outsource?
We may decide to outsource certain marketing functions that require specialized skills, are time-intensive, or are not needed on a continuous basis. Here are some functions that non-profits are most likely to outsource:
Who are the two people on your team that have the strongest influence over the decision of which marketing agency to hire?
In the context of making a strategic decision like hiring a marketing agency, the two individuals who would likely have the strongest influence on the decision-making process, apart from me as the Marketing Director, would probably be:
These individuals would likely have the strongest influence due to their strategic roles within the organization and their direct involvement with the financial and marketing aspects, respectively, of the decision to hire a marketing agency.
Now that you know the influencers of the decision, you can confirm that you’ve addressed (or at least acknowledged) their concerns. Or create content marketing assets specifically for them.
In this case, it could be a guide that explains how the ROI of social media fundraising campaigns is measured. The sales reps could send this to the finance director after a call or meeting.
Ardath is right. What do we do with all this stuff?
-Ardath Albee, Marketing Interactions
“A persona is simply a format that makes your customer insights useful.”
You found some insights, you’ve taken the time to validate them by talking to actual humans in your actual audience, then you edited the AI persona to improve the accuracy. Now what? How do we make this useful?
Without discussing the societal and labor market impacts of AI-generated content, I’ll briefly give a few use cases for applying the buyer persona.
Using the words that Taylor would use to describe his challenges and solutions, write a webpage promoting the social media marketing services for healthcare-focused nonprofits working on planned giving campaigns. Highlight the specific information that Taylor needs to know about an agency before hiring one.
I will not show the response. If I did, we may get distracted by this specific example and miss the point, which is that AI can indeed produce a workable first draft if given detailed prompts.
In this case, the generated text was 100% aligned with the persona. It basically wrote a section for each of the decision criteria we named above. It felt prescriptive because that’s exactly what it was.
This is an issue with prompting a generative AI. It’s a Goldilocks problem.
Mark O’Brien, Newfangled
“We’ve noticed something strange and surprising during our content experimentations and R&D efforts that seem to be part of the ‘nature’ of the ChatGPT models: it needs space to breathe. The more detailed and voluminous the instruction, the worse the output. The better and more succinct the prompt, the more creative, interesting, and natural the resulting language was.”
Also, the draft it generated lacked most of our 13 best practices for service pages, including these big ones:
I could have kept prompting it, training it to write a better first draft of a prospect-focused service page.
But the process of constructing that page will probably always be a mostly manual process. I don’t expect an AI to do an amazing job of gathering the right testimonials, curating the most relevant examples and results, creating branded visuals, etc.
It’s easy to imagine how this could be used to create ideas for content marketing or to validate advertising campaigns. We’ve focused on a B2B buyer persona, but AI could also help you understand B2C potential customers.
If the idea of AI generated webcopy leaves a bad taste in your mouth, here’s an idea that should cleanse your palette. Use generative AI to do gap analysis on an existing page, checking the copywriting against the AI marketing persona we worked so hard on these last 10 minutes.
The prompt is simple. Just ask which of the personas top concerns were not addressed in a given piece of web copy. You can simply copy and paste that copy into the prompt.
For this example, I found a random service page from a social media marketing company that works with non-profits.
Which of Taylor’s top concerns are not addressed on this webpage? [copy and paste in a webpage]
While the webpage addresses a number of general concerns related to non-profit marketing, there are some specific areas related to Taylor’s top concerns that are not explicitly addressed:
In essence, Taylor would likely appreciate more specificity related to his healthcare focus, his planned giving campaigns, cost-effectiveness, and value alignment from an agency.
You can see that it takes the task very literally and basically checks the page against the decision criteria we set in the earlier prompts and responses.
In this example, the page had a lot of misalignment because it wasn’t written for our persona. It was a random page. But you can see how AI is very useful at identifying gaps. A copywriter could use it as a quick double-check to confirm that nothing big was missed.
Even a single question left unanswered, or a single objection left unaddressed, can hurt conversion rates, as measured in Google Analytics.
Last step. Once you’ve got your new synthetic prospect in shape, make sure that you hold onto it. Name the chat so you can find it again quickly later.
All of your subsequent prompting about this target audience will take place in this chat. You can talk to it on a moment’s notice, day or night, for quick ideas or a long conversation about drivers, goals and motivations.
Never trust it completely.
Once you find insights, focus on execution.
What are your thoughts?