How to track and outperform competitors in AI assistant recommendations.
Understanding how AI assistants position your competitors is crucial for staying competitive. This guide shows you how to use PromptFern for competitive intelligence.
When users ask AI assistants for recommendations, they often get a short list of options. If your competitors consistently appear and you don't, you're losing opportunities before users even reach your website.
PromptFern helps you:
Navigate to Competitors in PromptFern and add:
For each competitor, capture:
Set up monitors specifically for competitive prompts:
Compare [your brand] vs [competitor A]
[Your brand] or [competitor B], which is better?
[Competitor A] vs [competitor B] vs [your brand]
Best alternatives to [competitor A]
[Competitor B] alternatives
Switch from [competitor A] to what?
Best [category] tools 2025
Top [category] software ranked
[Category] market leaders
Run your competitive monitors immediately to establish a baseline. You'll see:
Share of voice measures how often you're mentioned compared to competitors across all monitored prompts.
Calculation: Your mentions ÷ Total category mentions × 100
Example:
Your share of voice: 45 ÷ 160 × 100 = 28%
Track where you appear in recommendation lists:
| Prompt | Position 1 | Position 2 | Position 3 |
|---|---|---|---|
| "Best CRM tools" | Competitor A | Your Brand | Competitor B |
| "CRM for startups" | Your Brand | Competitor C | Competitor A |
| "Enterprise CRM" | Competitor A | Competitor B | Not mentioned |
Beyond mentions, understand the sentiment:
Compare your sentiment to competitors across the same prompts.
Look for prompts where competitors are weak:
Create content and messaging targeting these gaps.
Track when AI models criticize competitors:
Limitations of [competitor]
Problems with [competitor]
Why not to use [competitor]
If AI models highlight competitor weaknesses, ensure you address those concerns in your positioning.
In the Citations view, compare which sources cite you vs competitors:
If a competitor is heavily cited by certain publications:
If competitors lack content on specific topics:
When AI models compare you to competitors:
If competitors have more/better reviews:
Track week-over-week changes:
Deeper analysis:
Big-picture analysis:
Set up alerts for when new brands start appearing in your category responses.
When a competitor publishes new content, monitor whether it affects their AI visibility.
Try different phrasings in your content and see which generates better AI responses.
AI responses may vary by country or language. Track competitive position in each key market.
Competitive intelligence in AI is an ongoing process. Use PromptFern to:
Regular monitoring ensures you catch competitive movements early and can respond effectively.
I often do this thing where list items have headings.
For some reason I think this looks cool which is unfortunate because it's pretty annoying to get the styles right.
I often have two or three paragraphs in these list items, too, so the hard part is getting the spacing between the paragraphs, list item heading, and separate list items to all make sense. Pretty tough honestly, you could make a strong argument that you just shouldn't write this way.
Since this is a list, I need at least two items.
I explained what I'm doing already in the previous list item, but a list wouldn't be a list if it only had one item, and we really want this to look realistic. That's why I've added this second list item so I actually have something to look at when writing the styles.
It's not a bad idea to add a third item either.
I think it probably would've been fine to just use two items but three is definitely not worse, and since I seem to be having no trouble making up arbitrary things to type, I might as well include it.
After this sort of list I usually have a closing statement or paragraph, because it kinda looks weird jumping right to a heading.
I think most people are going to use highlight.js or Prism or something if they want to style their code blocks but it wouldn't hurt to make them look okay out of the box, even with no syntax highlighting.
Here's what a default tailwind.config.js file looks like at the time of writing:
module.exports = {
purge: [],
theme: {
extend: {},
},
variants: {},
plugins: [],
};Hopefully that looks good enough to you.
Nested lists basically always look bad which is why editors like Medium don't even let you do it, but I guess since some of you goofballs are going to do it we have to carry the burden of at least making it work.
The most annoying thing about lists in Markdown is that <li> elements aren't given a child <p> tag unless there are multiple paragraphs in the list item. That means I have to worry about styling that annoying situation too.
For example, here's another nested list.
But this time with a second paragraph.
<p> tagsBut in this second top-level list item, they will.
This is especially annoying because of the spacing on this paragraph.
As you can see here, because I've added a second line, this list item now has a <p> tag.
This is the second line I'm talking about by the way.
Finally here's another list item so it's more like a list.
A closing list item, but with no nested list, because why not?
And finally a sentence to close off this section.
I almost forgot to mention links, like this link to the Tailwind CSS website. We almost made them blue but that's so yesterday, so we went with dark gray, feels edgier.
We even included table styles, check it out:
| Wrestler | Origin | Finisher |
|---|---|---|
| Bret "The Hitman" Hart | Calgary, AB | Sharpshooter |
| Stone Cold Steve Austin | Austin, TX | Stone Cold Stunner |
| Randy Savage | Sarasota, FL | Elbow Drop |
| Vader | Boulder, CO | Vader Bomb |
| Razor Ramon | Chuluota, FL | Razor's Edge |
We also need to make sure inline code looks good, like if I wanted to talk about <span> elements or tell you the good news about @tailwindcss/typography.
code in headingsEven though it's probably a bad idea, and historically I've had a hard time making it look good. This "wrap the code blocks in backticks" trick works pretty well though really.
Another thing I've done in the past is put a code tag inside of a link, like if I wanted to tell you about the tailwindcss/docs repository. I don't love that there is an underline below the backticks but it is absolutely not worth the madness it would require to avoid it.
h4 yetBut now we have. Please don't use h5 or h6 in your content, Medium only supports two heading levels for a reason, you animals. I honestly considered using a before pseudo-element to scream at you if you use an h5 or h6.
We don't style them at all out of the box because h4 elements are already so small that they are the same size as the body copy. What are we supposed to do with an h5, make it smaller than the body copy? No thanks.
h4 elements, either.Phew, with any luck we have styled the headings above this text and they look pretty good.
Let's add a closing paragraph here so things end with a decently sized block of text. I can't explain why I want things to end that way but I have to assume it's because I think things will look weird or unbalanced if there is a heading too close to the end of the document.
What I've written here is probably long enough, but adding this final sentence can't hurt.
I've also added support for GitHub Flavored Mardown using remark-gfm.
With remark-gfm, we get a few extra features in our markdown. Example: autolink literals.
A link like www.example.com or https://example.com would automatically be converted into an a tag.
This works for email links too: contact@example.com.