We have built multiple ways of generating prompts:
You can upload questions.
You can convert your Google Search Console data to prompts.
You can use our AI composer to generate prompts.
In our AI Composer we have lots of powerful predefined nodes, such as question types, expertise levels, buyer journey, buyer personas, etc., to help you get started quickly.
But sometimes you want to really customise the prompt set to exactly describe what you're looking for and what you're not looking for.
By default, our prompts are unbranded.
If you want to produce a set of branded prompts to see how large language models talk about your brand when your brand is actually mentioned in a prompt then you need to use our Custom Modifier.
This node allows you to be extremely creative, just as if you were using a real AI model to create an image, so you're only limited by your imagination.
Using Custom Modifiers for Prompt Engineering
Simply add a custom modifier node to the canvas and edit it. You can use it for many things such as:
Creating prompts in specific languages
Creating prompts aimed at consumers from specific countries or locales
Creating A/B style market research questions to explore bias
Asking ranking or rating-style questions to get LLMs to rate you versus your competitors
Generating reputational management questions by exploring negatively framed prompt sets
In all of these examples, you can just use a custom modifier. This allows you to put in a completely custom prompt, which our prompt generation engine will use with the rest of your flow to generate exactly the type of prompts that you want.
Do bear in mind we don't charge for prompt generation, so you can iterate and explore until you get exactly the right type of prompts you want.
Here's an example of the empty node. Just add it to either the brand node or a category node, depending on where you want these types of questions to appear. Typically, you'd add it right at the top of the flow at the brand level.
Obviously, we recommend separating branded queries or competitive head-to-heads or anything with a brand term from your unbranded queries because this will skew the results.
Once you have added the Custom Modifier to your flow. You can simply edit it. Here's an example for adding a custom location, but you could also use this for languages or any of the custom prompts that are set out below.
Creating Branded Prompt Sets
Our default position is that we generate unbranded prompts, so to generate branded prompts, just add a custom modifier to your flow.
Example: this prompt creates branded questions about Ford.
Questions about Ford
Generate a representative list of natural-language consumer questions that someone in the United States might ask when considering purchasing a Ford vehicle. The questions should span the entire buyer journey — from early-stage discovery to evaluation, comparison, financial considerations, and ownership concerns. Every question must explicitly mention the Ford brand or a specific current Ford model (e.g., F-150, Escape, Mustang Mach-E, Explorer, Maverick). Use diverse phrasing and reflect real-world buyer intent. Group questions into the following stages of the buyer journey: Awareness, Consideration, Evaluation, Purchase, and Post-Purchase.
Awareness “Is Ford a reliable car brand in 2025?” “What are the newest Ford SUVs available in the U.S.?”
Consideration “How does the Ford Escape compare to the Toyota RAV4 for fuel economy?” “Is the Mustang Mach-E a good option for families?”
Evaluation “What are the most common problems with the Ford F-150?” “Are Ford hybrids better than Honda hybrids?”
Purchase “What’s the best time of year to buy a Ford Explorer?” “Does Ford offer 0% financing on the Maverick?”
Post-Purchase “How much does it cost to service a Ford Bronco?” “Is Ford’s extended warranty worth it?”
Ensure that every generated question MUST include Ford or a specific current Ford model name.
[The last line telling the model that it MUST include Ford is very important].
Creating Head-to-Head Prompt Sets vs Competitors
Just like real consumers do, large language models have biases and preferences for certain brands. You can explore this by creating a custom prompt set with AB-style questions asking the large language model to compare you against a brand across different criteria, categories, and key factors across all your flows.
Prompt Set Comparing Gousto vs Hello Fresh
Gousto vs hellofresh
Generate comparative questions that pit HelloFresh and Gousto against each other in the UK meal-kit delivery market. Each question should follow this format: “Which meal-kit delivery service should I [action]?” a) HelloFresh or b) Gousto? Only answer A or B. For every comparison, duplicate the question and reverse the order of options A and B, so one version reads a) HelloFresh or b) Gousto, and the other reads a) Gousto or b) HelloFresh. This controls for ordering bias when evaluating responses from language models. Guidelines: Focus exclusively on HelloFresh and Gousto — no other brands. Use natural, UK consumer phrasing (e.g. “recipe boxes”, “meal plans”, “portions”, “delivery days”, “postcode coverage”).
Make questions practical and relatable to real-world decision-making. Cover all key aspects important to UK meal-kit buyers, including: freshness • taste • recipe variety • portion size • dietary options • healthiness • price • value • ease of use • delivery • packaging • sustainability • flexibility • customer service • overall satisfaction. Ensure both brands appear equally viable and well-known — avoid biased wording or clear “winners”.
Output only the questions, not the answers. Example Outputs Meal quality and variety Which meal-kit delivery service should I try for fresher ingredients? a) HelloFresh or b) Gousto? Only answer A or B. Which meal-kit delivery service should I try for fresher ingredients? a) Gousto or b) HelloFresh? Only answer A or B. Ease and convenience Which meal-kit delivery service should I use for quicker weekday dinners? a) HelloFresh or b) Gousto? Only answer A or B. Which meal-kit delivery service should I use for quicker weekday dinners? a) Gousto or b) HelloFresh? Only answer A or B. Cost and value Which meal-kit delivery service should I choose for better value per portion? a) HelloFresh or b) Gousto? Only answer A or B. Which meal-kit delivery service should I choose for better value per portion? a) Gousto or b) HelloFresh? Only answer A or B. Sustainability and packaging Which meal-kit delivery service should I pick for more eco-friendly packaging? a) HelloFresh or b) Gousto? Only answer A or B. Which meal-kit delivery service should I pick for more eco-friendly packaging? a) Gousto or b) HelloFresh? Only answer A or B. Customer experience Which meal-kit delivery service should I trust for better customer support? a) HelloFresh or b) Gousto? Only answer A or B. Which meal-kit delivery service should I trust for better customer support? a) Gousto or b) HelloFresh? Only answer A or B.
Creating Prompt Sets for Proactive Reputation Management
If you're interested in understanding all the bad stuff that large language models and AI answer engines might say to real users about you and your competitors, then just create a negatively framed prompt set. Run these questions, and you can see all the sources the LLM cite when saying detrimental things about your brand! Here's an example for UK airlines.
Negative Bias
Generate a list of standalone, unbranded, negatively biased search questions that real consumers might ask about when choosing flights and airlines. Focus on common frustrations, risks, or concerns related to air travel, including topics like delays, cancellations, pricing, hidden fees, environmental impact, comfort, customer service, and safety. Each question should be phrased naturally, as if typed into a search engine or asked to an AI assistant. Avoid naming specific airlines, but make the issues realistic and relatable for travelers and frame the questions negatively.e.g Which airlines should I avoid? Which airlines are the least safe? Which airline has the worst customer service?
If you want to get started really quickly, we recommend taking one of these and using it as a starting point, then getting ChatGPT or Claude to edit it for you for your specific brand or use case.