Improving Your AI Visibility - A Complete Guide

AI assistants are becoming a primary way people discover and evaluate brands. This guide explains how to improve your brand's presence in AI responses.

Understanding AI Visibility

When someone asks ChatGPT "What's the best CRM for startups?" or Perplexity "Compare Notion vs Coda", the AI's response is influenced by:

1. Training data — Content the model learned from during training
2. Retrieval sources — Websites the model accesses in real-time (Perplexity, AI Overviews)
3. User context — Previous conversation and stated preferences

Your goal is to influence these factors so AI models accurately represent your brand.

Step 1: Audit Your Current Visibility

Before optimizing, understand where you stand.

Use PromptFern to Monitor

Set up monitors for key prompts in your category:

• Discovery prompts: "Best [category] tools"
• Comparison prompts: "Compare [your brand] vs [competitor]"
• Recommendation prompts: "What [category] should I use for [use case]?"

Analyze the Responses

Look for patterns:

• Are you mentioned at all?
• Where do you appear in recommendation lists?
• What do AI models say about you (positive, neutral, negative)?
• Which competitors appear alongside or instead of you?

Step 2: Create AI-Optimized Content

AI models learn from content across the web. Create content that directly answers the questions users ask.

Target Common Questions

For each key prompt you're monitoring, create content that answers it:

Structure Content for AI

AI models process structured content better:

• Use clear headings (H2, H3) that match question intent
• Answer questions directly in the first paragraph
• Use lists and tables for easy parsing
• Include relevant keywords naturally
• Keep content updated — fresh content gets indexed

Example: Comparison Page Structure

# [Your Brand] vs [Competitor] - 2025 Comparison
 
## Quick Summary
[Direct answer to which is better for what use cases]
 
## Key Differences
 
| Feature | Your Brand | Competitor |
|---------|-----------|------------|
| Pricing | $X/mo | $Y/mo |
| Feature A | ✓ | ✗ |
| Feature B | ✓ | ✓ |
 
## Who Should Use [Your Brand]?
[Clear recommendation for your target audience]
 
## Who Should Use [Competitor]?
[Honest assessment of when competitor is better]
 
## Conclusion
[Balanced summary with clear recommendation]

Step 3: Build Authority Signals

AI models weight sources by authority. Get featured in sources AI trusts.

Target These Source Types

1. Review Platforms

   • G2, Capterra, TrustRadius
   • Encourage customers to leave reviews
   • Respond to all reviews professionally

2. Industry Publications

   • Guest posts on relevant blogs
   • Interviews with trade publications
   • Inclusion in "best of" roundups

3. Comparison Sites

   • Ensure listings are accurate
   • Provide case studies and data

4. Wikipedia (if notable)

   • Factual, well-sourced entries
   • Follow notability guidelines

Citation Strategy

For AI models with real-time retrieval (Perplexity, AI Overviews):

• Monitor which sources they cite for your category
• Target those sources for content or mentions
• Track citation trends in PromptFern

Step 4: Technical Optimization

Help AI crawlers understand your content.

Allow AI Bots

Check your robots.txt:

User-agent: GPTBot
Allow: /

User-agent: Google-Extended
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: anthropic-ai
Allow: /

Implement Schema Markup

Help AI understand your content structure:

{
  "@context": "https://schema.org",
  "@type": "SoftwareApplication",
  "name": "Your Product Name",
  "description": "What your product does",
  "applicationCategory": "BusinessApplication",
  "operatingSystem": "Web",
  "offers": {
    "@type": "Offer",
    "price": "99",
    "priceCurrency": "USD"
  }
}

Monitor AI Bot Traffic

Use PromptFern's Website Analytics to:

• Verify AI bots are crawling your site
• See which pages they access
• Track crawl frequency

Step 5: Monitor and Iterate

AI visibility is not set-and-forget. Models update regularly.

Weekly Review

• Check PromptFern for visibility changes
• Review new responses mentioning your brand
• Track competitor movements

Monthly Analysis

• Analyze citation sources for opportunities
• Identify content gaps
• Update existing content with fresh information

Quarterly Strategy

• Review overall visibility trends
• Adjust monitoring prompts based on market changes
• Evaluate competitive positioning

Common Mistakes to Avoid

1. Keyword Stuffing

AI models detect and penalize unnatural content. Write for humans first.

2. Ignoring Competitors

If competitors consistently outrank you, study what they're doing differently.

3. Outdated Information

AI models may learn outdated information. Keep content current.

4. Missing the Question

If your content doesn't directly answer the questions people ask, AI won't reference it.

5. Neglecting Reviews

Customer reviews heavily influence AI recommendations. Actively gather and respond to them.

Measuring Success

Track these metrics in PromptFern:

Conclusion

Improving AI visibility is a long-term effort combining content strategy, authority building, and continuous monitoring. Use PromptFern to track your progress and identify opportunities for improvement.

• 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.

Code should look okay by default.

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.

What about nested lists?

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.

1. Nested lists are rarely a good idea.
   • You might feel like you are being really "organized" or something but you are just creating a gross shape on the screen that is hard to read.
   • Nested navigation in UIs is a bad idea too, keep things as flat as possible.
   • Nesting tons of folders in your source code is also not helpful.
2. Since we need to have more items, here's another one.
   • I'm not sure if we'll bother styling more than two levels deep.
   • Two is already too much, three is guaranteed to be a bad idea.
   • If you nest four levels deep you belong in prison.
3. Two items isn't really a list, three is good though.
   • Again please don't nest lists if you want people to actually read your content.
   • Nobody wants to look at this.
   • I'm upset that we even have to bother styling this.

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.

  • These list items won't have <p> tags
  • Because they are only one line each

• But 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.

There are other elements we need to style

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:

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.

Sometimes I even use code in headings

Even 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.

We haven't used an h4 yet

But 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.

We still need to think about stacked headings though.

Let's make sure we don't screw that up with 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.

GitHub Flavored Markdown

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.