Before you choose: what a good prompt looks like
Think buyer language, not internal language. A prompt is a question or query someone would type into ChatGPT, Perplexity, or Gemini when looking for something in your category. Good prompts are:- Questions your buyers actually ask - not how you describe your product, but how they describe their problem. “What’s the best platform for tracking AI search visibility” beats “AthenaHQ competitors.”
- A mix of Branded and Non-Branded - Branded prompts include your company or product name. Non-Branded prompts are generic category queries where you want to show up without being named. You need both.
- Specific enough to return signal - very broad prompts (“what is marketing”) return noise. Specific ones (“what tools track brand mentions in AI responses”) return the data you can act on.
Option 1 - Add Prompts directly
This is the quickest path for getting prompts into Athena. You open the Add Prompts drawer and choose how you want to add them: let Athena generate suggestions, type them in yourself, or upload a CSV. Access it from: Prompts > + Add More > Add Prompts Once the drawer opens, you’ll see three ways to add:AI generation
Athena generates a starting set of prompts for you based on your website, brand description, and industry. This is the fastest way to get something in place if you’re starting from scratch and want Athena to do the thinking. Best for: New customers who haven’t done keyword research yet and want a reasonable starting point immediately. Good for filling coverage gaps when you know a topic area you’re missing but aren’t sure what the right prompts are.Manual entry
Type or paste prompts directly into the drawer. You set the type (Branded or Non-Branded), assign a topic, choose a country, and add tags before saving. Best for: When you know exactly what you want to track. Common use cases: prompts you heard from customers, questions from sales calls, terms your team competes on, or specific product comparisons you want to monitor. This is also the right move when AI generation misses something obvious - just add it yourself.CSV import
Upload a formatted CSV to add prompts in bulk. Download the template from within the drawer, fill it out, and drag the file to upload. If you attach locations to prompts, separate multiple locations with semicolons (London; Manchester). Commas are treated as part of the location name, so a full address like Acme Autocentre, 12 Windsor Way, Aldershot, GU11 1JG stays one location. In the Country column, list one country per location in the same order, also separated by semicolons.
Best for: Teams coming from an SEO tool with an existing keyword export, or any situation where you have a list of 20+ prompts you want to load in one go rather than one by one. Also useful for migrating prompt sets from another system, or when multiple stakeholders have contributed prompts to a shared spreadsheet.
Tips for all three:
- After adding, review the Topics Athena assigns - the defaults are a starting point, not final. Drag prompts between topics to get your structure right, since Topics are how you’ll slice data in Olympus and the Heatmap.
- Once a prompt has collected response data, its text becomes read-only. You can’t edit the wording - you’d need to pause it and create a new one. Get the phrasing right upfront.
Option 2 - Discover
Open Prompts > Discover > Run discovery. Access depends on your plan and the sources available to your website.- Name the run and choose a country and language.
- Select Website analysis, Search Console, Social, Keyword gaps, or Custom instructions. Search Console requires a connected property. Custom instructions requires a written brief.
- Run discovery, then open the run from history to follow progress and check the source summary.
- Review the suggested prompts by topic. Filter by Source, Funnel, Intent, or Type, and compare the available Volume and Value estimates.
- Select the prompts you want and click Start tracking. Review the new and already-tracked counts before confirming.
Choosing the right option
These methods layer well together. A common starting pattern: use AI generation or CSV to get something tracking immediately, then run discovery to fill structural gaps and catch competitive exposure you didn’t know to look for.
After you add prompts: organizing and managing them
Once prompts are in, a little structure makes your data significantly more useful. Topics are the main organizational unit. Prompts are grouped into Topics on the Manage tab. Think of Topics like content pillars or product areas - “Pricing,” “Competitors,” “Use Cases,” and so on. A well-organized topic structure makes the Heatmap and Olympus filter bar much more powerful, since you can slice all your metrics by Topic to see exactly where you’re winning or losing. To move prompts between Topics, select them and use More > Assign topic in the bulk command bar, or drag them directly onto a different topic row. Branded vs. Non-Branded is the other key label. Branded prompts include your company or product name. Non-Branded are generic category queries. Track both - Branded mention rate reflects direct awareness; Non-Branded reflects whether you’re showing up when buyers haven’t decided yet. You can set or edit this in the Edit Prompt drawer (click … > Edit on any prompt row). Tags let you create custom groupings across Topics - useful for things like campaign tracking, product lines, or funnel stage labels you want to apply manually. Manage the full tag library from + Add More > Add Tag. Pausing prompts is the right move for anything off-target - don’t delete, since paused prompts keep their historical data. Select the rows and click Pause in the bulk command bar, or use the row-level … menu. A note on editing prompt text: once a prompt has collected response data, its text is locked. You can’t edit the wording directly - the Edit Prompt drawer will have the text field disabled. If you need to change it, pause the prompt and create a new one with the corrected text. Plan your phrasing carefully upfront and you won’t need to do this often.Set up recurring schedules
Open Schedules from the sidebar. Each schedule has its own prompt selection, models, frequency, and run options. You can edit Default or click New schedule to create a separate recurring run.- Name a new schedule and choose All active prompts, or add conditions by prompt tag, topic, prompt type, or country. Every condition must hold. Within a condition, a prompt matches any of the chosen values, and a condition can be set to “is not any of” to leave prompts out. A prompt matched by several values is included once.
- Choose the AI models to query.
- Set a weekly cadence, including the selected days and a repeat interval of one to four weeks, or choose a monthly day.
- Review Run options, including Fan-out, Personas, All assigned locations, Additional personas, and Additional locations.
- Check the run preview and estimated credits, then save the schedule.