Playbook · methodology
Win comparison prompts in your category
When buyers ask AI “what's the best … for …?” or “A or B?”, the answer is a shortlist. This playbook measures whether you are on it, finds out why competitors are, and shows what to change.
- Goal
- Higher mention rate and share of voice in comparison answers
- For
- Brands in competitive categories, SaaS, e-commerce
- Baseline
- 7 days of daily tracking
- Re-measure
- After 6–8 weeks, then monthly
Goal
Increase how often AI engines name your brand, and how prominently, when people ask for recommendations or comparisons in your category. The primary metrics for a fixed prompt set are mention rate (the share of answers that mention you) and share of voice (your mentions relative to those of the competitors you track). Position and sentiment are secondary.
Who it's for
- Brands in categories where buyers compare options before they buy
- SaaS and software companies competing on “best tool” lists
- E-commerce brands competing for product recommendations
- Agencies preparing a competitive GEO analysis for a client
Prompt set
Twelve templates covering the three shapes comparison questions take: shortlist (“best”), head-to-head (“vs.”) and alternative (“instead of”). The tags let you filter results by shape later.
Import it in AutoSEO under AI Visibility → Tracker → Import CSV (paste the text or upload the file).
Replace these placeholders before you import — the importer takes every line literally:
[category]— your product category in your buyers' words, e.g. “project management software”[audience]— a buyer segment, e.g. “small agencies”[use case]— a job to be done, e.g. “client reporting”[competitor]— your strongest competitor (duplicate the rows for more)[brand]— your brand name[budget]— a realistic constraint, e.g. “under $50 a month”
prompt,tags"What is the best [category] for [audience]?","Comparison|Shortlist""Which [category] do experts recommend?","Comparison|Shortlist""Top 5 [category] for [use case]","Comparison|Shortlist""What is the best [category] [budget]?","Comparison|Shortlist|Budget""Which [category] is easiest to set up for [audience]?","Comparison|Shortlist""[brand] vs. [competitor]: which is better for [use case]?","Comparison|Head-to-head""[competitor] vs. [brand] for [audience]","Comparison|Head-to-head""Is [brand] better than [competitor]?","Comparison|Head-to-head""What are the best alternatives to [competitor]?","Comparison|Alternative""What should I use instead of [competitor] for [use case]?","Comparison|Alternative""Cheaper alternative to [competitor] for [audience]","Comparison|Alternative|Budget""What are the pros and cons of [brand]?","Comparison|Brand"Engines to track
Comparison answers differ a lot between engines, so track several — and keep the same engines for the whole measurement period.
| Engine | Why it matters for comparison prompts |
|---|---|
| ChatGPT | The assistant most people use for open questions; with web search, it cites the sources behind its shortlist. |
| Perplexity | Cites sources in every answer, so you see exactly which comparison pages shape the shortlist. |
| Google AI Overviews and AI Mode | Appear above or instead of classic results for “best …” searches. |
| Gemini, Claude, Microsoft Copilot | Add them if your buyers work in Google Workspace, Claude or Microsoft 365. |
Baseline measurement in AutoSEO
- 1
Create the project
Add your website as a project, set the country and language of your target market and add three to five competitors under AI Visibility → Competitors.
- 2
Import the prompt set
Replace the placeholders, then import the CSV under AI Visibility → Tracker → Import CSV. The tags are imported with the prompts.
- 3
Choose engines and frequency
Under AI Visibility → Model Settings, select the engines from the table above and set tracking to daily for the baseline week.
- 4
Freeze the baseline
After seven daily runs, note mention rate, share of voice, position and sentiment per tag, and save a report from the Baseline Snapshot template.
Actions
- 01
Read the answers, not just the scores
Open the answers that name competitors but not you. Note the reasons AI gives — price, integrations, audience fit. Those reasons are your content brief.
- 02
Fix the sources behind the shortlist
AI Visibility → Sources shows which pages and domains are cited for these prompts. Work on being included where you are missing: comparison articles, review platforms, directories.
- 03
Publish honest comparison pages
Create “[brand] vs. [competitor]” and “alternatives to [competitor]” pages with accurate, dated facts, a clear verdict per use case and a comparison table.
- 04
Make key facts easy to extract
State pricing, target audience, integrations and limits as plain text on crawlable pages, not only in images, PDFs or scripts. Check it with Crawlability.
- 05
Work through the task list
Optimizations → Tasks turns tracker, source and audit evidence into prioritized tasks. Push them to Jira, Linear or Asana so they actually get done.
- 06
Keep the prompt set frozen
Don't edit or remove prompts during a measurement period. Add new prompts under a new tag so the original set stays comparable.
How to measure change
- Compare the same prompt set, engines, country and language as in the baseline
- Use averages over at least seven runs per period, never single days
- Look at mention rate and share of voice per tag (shortlist, head-to-head, alternative) separately
- Check whether the sources you worked on now appear under Sources for these prompts
- Watch sentiment and the reasons AI gives, not only whether you are mentioned
- Present before and after with the Competitor Benchmark or Monthly Report template
A simple decision rule
Treat a change as real only if it holds for two consecutive measurement periods and shows up in more than one engine. A one-week spike in a single engine is usually noise.
Expected variance
AI answers are probabilistic. The same prompt can produce a different shortlist tomorrow, in another country or after a model update. Expect mention rates to move by several percentage points from day to day even when nothing changed on your side — the smaller the prompt set, the larger the swings.
Three habits make your numbers more trustworthy:
- More prompts: twelve prompts give a rough signal; 30–50 prompts per topic give a much steadier one.
- More runs: daily tracking over a week averages out the randomness of single answers.
- A stable setup: changing engines, providers, country or language breaks comparability. AutoSEO labels simulated engines, so you can keep them apart from native answers.
Limitations
- Tracking shows what engines answer to your prompts, not how many real people ask them — there is no equivalent of search volume for prompts.
- Answers retrieved through APIs or data providers can differ from what a signed-in user sees in a consumer app with personalization and memory.
- Changes to your pages only take effect after engines recrawl or retrieve them again; the timing is outside your control.
- Share of voice is relative to the competitors you track. Changing the competitor list changes the comparison, so keep it fixed for a before-and-after.
Questions about this playbook
How many prompts do I need?
Start with the twelve templates, expanded to your categories and competitors. For decisions, aim for 30 to 50 prompts per topic, because more prompts smooth out the randomness of individual answers. Keep the set fixed during a measurement period.
How long until I see a change?
There is no reliable timeline. Engines that search the web live, such as ChatGPT with search, Perplexity and Google AI Overviews, can pick up new or updated pages once they are crawled. Knowledge from model training only changes with new model versions. Re-measure after six to eight weeks and then monthly.
Should my prompts include my brand name?
Use both kinds and tag them separately, because they measure different things. Unbranded prompts such as “best [category]” show whether AI recommends you without being asked about you, which is the main goal here. Branded prompts such as “[brand] vs. [competitor]” show how AI judges you once you are part of the question.
Is it fair to publish comparison pages about competitors?
Yes, if they are accurate, current and fair. Misleading comparisons damage trust and can breach comparative-advertising rules in many countries. Cite your sources, date your facts and update the pages when products change.
Can I run this playbook for a client?
Yes. Create a project for the client, import the prompt set and use the Pitch or Competitor Benchmark report templates for the presentation. Give the client read-only access with the Client role.
Run this playbook in AutoSEO
Every feature is included in the free self-hosted edition and in AutoSEO Cloud.