How to get ChatGPT to recommend your brand

Ask ChatGPT for "the best life insurance for self-employed people" and look at which brands it names. If you sell exactly that and you're missing, the model almost never ignored you on purpose. It just couldn't find a sentence of yours it was able to use. This guide covers how to get ChatGPT to recommend your brand, based on a Spanish-language webinar we ran in August 2025 with SEO consultant Ana Fernández, who has spent more than 10 years in content marketing and SEO, and on how we measure brand visibility at SearchBrand.

One caveat up front. The session was recorded on August 25, 2025. Since then Google AI Mode, ChatGPT search and our own platform have all changed. What follows keeps the advice that still holds a year later and drops every figure that was shared without a source.

Why ChatGPT doesn't recommend your brand

Ana's starting point, at 22:14 in the video: models don't make information up. They look it up, blend it and hand it back in a form that's easy to read. When someone asks about a category, ChatGPT builds the answer from what it learned in training plus what it finds on the web. If none of those sources says plainly what you are and who you serve, the model recommends a brand it can describe.

The webinar pointed to three causes.

Your homepage talks about "we"

"We are leaders in protection solutions for you and your family." Some version of that line sits on thousands of insurance sites. A visitor already on your page gets it. A model reading that one sentence on its own finds no brand name, no category and no location.

Ana explained why: algorithms read each sentence as a unit and don't assume "we" means the company that owns the domain. Her example was a Chilean newspaper story about a Trump–Zelensky meeting. If a model tied every sentence to the site that published it, it would think the paper had something to do with the meeting.

"If you're not stating it explicitly, AI won't be able to work it out for you." — Ana Fernández, SEO consultant [22:14]

Nobody else mentions you

The biggest gap versus classic SEO, in Ana's view, sits off your site. A model that wants to give reliable answers leans on third parties: a news story, a Google review, a Reddit thread. If your brand only exists on your own domain, you're one source against competitors with ten.

Your ChatGPT rival is someone else

Ana made this part of step one: your direct competitor on Google "won't necessarily be" your competitor on ChatGPT. The brand that beats you on page one for "life insurance" may not show up in ChatGPT at all, while a comparison site stuck on page two of Google gets named first. If you don't know who beats you on each channel, you'll copy the wrong rival.

ChatGPT AEO starts with measurement

"You can't optimize what you don't measure," Ana said as the practical part began. Most brands skip this step and go straight to rewriting. Answer engine optimization for ChatGPT only works if you know where you stand today and who's taking your spot.

Pick 10 prompts the way your buyer writes them

Short keywords are fading. Ana's example: people used to search "running shoes" and now type something like "what are the best running shoes if I run on asphalt at night and the ground is dry." Search volume per keyword matters less when traffic spreads across "thousands of unique micro-queries."

Build a list of 10 prompts in that shape:

• 4 category prompts ("best life insurance for freelancers in the US")
• 3 comparison prompts ("is a broker better than buying direct from an insurer?")
• 2 decision prompts ("is whole life worth it at 35 if I'm self-employed?")
• 1 brand prompt ("what do people say about Andes Life?")

Run them in ChatGPT and Gemini, and in Perplexity and Google AI Mode if you can, always from the market you care about. For each answer, log whether you appear, where, which competitors show up and which sources the model cites.

Share of voice tells you if you're winning

From that log you get share of voice: the percentage of answers that mention you. Show up in 6 of 40 answers (10 prompts across 4 models) and your share of voice is 15%. If your main rival shows up in 22, they're at 55% and the gap is 40 points. Re-run the same prompts every month. When your number drops while a competitor's climbs, you're losing ground in AI answers even if your Google traffic hasn't moved.

In the webinar demo, Hussam showed what this looks like with a sample Netflix account in Chile: position 1 in the ChatGPT answer, 9 brands mentioned, 4 sources cited and a 52% share of voice across the tracked questions. The product has changed a lot since August 2025, but the logic hasn't.

State your positioning in one sentence

This is the change that moves results most and costs least. Ana put it bluntly: if you're the best insurer in Peru, you have to say "I'm the best insurer in Peru." It feels odd to write that on a homepage. It's what the model needs to read.

The formula we use in audits has five parts:

Part	Andes Life (example)
Brand	Andes Life
Category	life insurance broker
Audience	freelancers and self-employed workers
Market	the US
Verifiable fact	compares 12 insurers, quote within 24 hours

Result: "Andes Life is a life insurance broker for freelancers and self-employed workers in the US that compares 12 insurers and sends your quote within 24 hours." That's 27 words and it stands on its own.

Put that sentence on your homepage and your product page. Then carry it to your external profiles: the LinkedIn bio, Instagram, your Google Business Profile. Ana stressed keeping the story "clear and cohesive" on every channel. When each profile describes you differently, the model ends up with five versions and none of them wins.

Write sentences a model can lift

"An extractable sentence is one where, if ChatGPT took it whole and dropped it into its chat, it would work." — Ana Fernández, SEO consultant [27:07]

Ana called this her favorite part. Her theory is that building answers costs models compute, and a ready-made sentence saves them some. That's her hypothesis, not something OpenAI has published. What you can check is whether your sentences have the traits she described, and the auditor below turns them into rules. Aim for 12 to 35 words. Name the brand next to the category. Don't open with "this" or "also", which force the reader back to the previous sentence.

Before and after, same facts:

• Before: "We've been helping our clients for over 7 years."
• After: "Andes Life has quoted life insurance for self-employed workers in the US for 7 years and placed more than 4,000 policies."

The second one names the brand and carries two numbers someone can check. The first has a number with no owner.

Use data only you have

The webinar flagged proprietary hard data as something AI responds to. In practice that means your renewal rate, your real turnaround time, your starting price. A figure nobody else has gives the model a reason to cite you instead of an aggregator.

Write for the question, not the keyword

Ana suggested writing around conversational prompts like "can I sign up for Talent Pro if I'm a small business?", with your own product name in place of the example Turn those questions into subheadings and answer each one in the first sentence below it. Her trick for structure: ask ChatGPT about your topic (not how to optimize for it) and look at how it organizes the answer. That layout hints at what it expects to find.

Cover the whole topic

The most technical stretch of the webinar was about semantic fields. Models learn which words sit near each other. Ana's example: a model has read "coffee with milk" thousands of times and "coffee with tires" never, so it finishes the phrase with milk. You want your brand to sit next to your category so often that it works the same way.

You get there two ways. One is the extractable sentences above. The other is covering the topic's whole neighborhood: if you sell life insurance, you need articles on reading a policy or deciding when a savings component pays off, linked to each other and to your product page. According to Ana, that's why Wikipedia ranks so well: it covers every topic and links everything.

On which formats get cited, Ana said that outside Wikipedia and Reddit, most citations in the studies she follows go to blog posts and help-center articles, with very few going to product pages. She didn't name the source, so treat it as a lead and test it against your own prompts. Her practical advice was to publish comparisons and lists ("the 10 best tools for…", "your brand vs. a rival"), since they hand the model tidy, short information it can reuse. We cover how to run that kind of check in our AEO and GEO audit guide.

What happens off your site

An attendee asked whether classic SEO was enough. Ana's split: 80% the same, 20% different. That 20% lives on other people's pages.

PR, in her words, is having "a revival." ChatGPT is a business that needs to give correct information, and citing sites it already trusts is one way to protect itself. News outlets have spent years earning that trust. Her priority list started with press mentions and interviews. Then come Reddit and industry forums, third-party reviews with real testimonials, and a message that doesn't change from one social profile to the next.

On backlinks she was candid: "I'm not particularly a fan of backlinks as such; I prefer press mentions." She dislikes the volume game and links that don't support any story. We agree and go further: we don't recommend buying link packages to show up in AI answers. One trade-press article that repeats your positioning sentence is worth more than 50 directory links.

The technical floor

Short section, because the list is short:

• Organization and FAQPage schema; Product or Review if you sell something specific. Ana said structured data always mattered to Google and now carries more weight.
• Main copy in the initial HTML. If your site needs JavaScript to show content, many AI crawlers see an empty page. Server-side rendering or prerendering fixes it.
• Nothing important hidden. In Ana's words, don't bury key information "behind 14 cards that flip over," because crawlers don't click.
• Clear bullets and subheadings, which help readers and crawlers alike.

What about video? Asked by an attendee, Ana said YouTube helps in Google, which shows video and Shorts blocks, but ChatGPT doesn't cite it much yet. AI Mode does value multimodal content. If you publish video, add captions: text can be crawled, the footage mostly can't.

Keep the human in it

Ana shared a case from a company that sells courses to people aiming for Wall Street jobs. Its articles on consulting-firm interview questions were among its top sellers, and they started slipping. The cause turned out to be an AI Overview answering the question before the click. The team changed only the title, to something like "the questions to get into this firm, from someone who's been through them." According to Ana, clicks "shot up" with the body untouched. She gave no numbers, and the lesson doesn't need them: people still want to hear from someone who's been there, and structuring for AI doesn't mean writing without a voice.

What we don't recommend

• Rewriting the whole site before measuring. Without your 10 prompts and a baseline share of voice, you won't know what worked.
• Filling the homepage with "we're the best" and no proof. Explicit positioning needs a number next to it so the model reads it as information, not advertising.
• Buying links or reviews. They add risk and don't change how you're described.
• Mass-producing AI articles for every keyword variant. A cluster of 4 solid, interlinked articles helps the model more than 40 near-duplicate pages.
• Trusting a single measurement. Answers vary from one model to another and from week to week; re-run the same prompts monthly.

Audit your page: can AI cite you?

The auditor checks your homepage or product page copy against the same rules we apply by hand. Paste the text and describe your brand in four short fields; optionally add your bios and your buyers' questions. Then tick an 8-item checklist of technical and off-site signals. You get a 0–100 score across three levels, the "we" sentences worth rewriting, three suggested extractable sentences, a schema block ready to paste and, most useful of all, the three fixes to make first. Everything runs in your browser and nothing you paste is stored.

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Andes Life",
  "description": "Andes Life is a life insurance broker for freelancers and self-employed workers in the US.",
  "alternateName": [
    "AndesLife"
  ],
  "url": "https://andeslife.example",
  "areaServed": "the US"
}

Worked example: rewriting Andes Life step by step

Andes Life is a fictional broker we use to show the full process. The numbers are illustrative.

1. Measure. The team ran 10 prompts across 4 models. Andes Life appeared in 3 of 40 answers (7.5% share of voice). The brand ChatGPT cited most was a comparison site sitting at position 14 in Google.
2. Audit the homepage. On the original copy ("We are leaders in protection solutions…") the auditor scored 6 out of 100: four "we" sentences with no brand, no category statement, zero data. Its first flag was key information hidden in a carousel.
3. Write the positioning sentence. "Andes Life is a life insurance broker for freelancers and self-employed workers in the US that compares 12 insurers and sends your quote within 24 hours."
4. Replicate it. The same sentence went onto the product page and the three external profiles (LinkedIn, Instagram, Google Business Profile).
5. Cover the semantic field. Four interlinked articles: how to read a policy, how to get a quote as a 1099 worker, term vs. whole life, and self-employed vs. employer coverage. Each with FAQ and schema.
6. Go off-site. One personal-finance press mention, a review request to the last 20 customers and answers in two freelancer forums.
7. Measure again. Live measurement repeats every 30 days for three months with the same 10 prompts. If share of voice is still flat at day 60, check which sources the model cites and get a presence on those.

Watch the webinar with Ana Fernández

The webinar is in Spanish and runs 1 hour 9 minutes. If you're short on time, start with key steps to appear in AI answers (22:14) and then human and technical content strategy (27:07). The PR and reputation section starts at 40:08. The closing demo shows SearchBrand as it was in August 2025; the pricing and offer mentioned there no longer apply.

Ana closed with a line that works as a filter for everything above:

"How do I make the experience so good for a user that the algorithm can't ignore it?" — Ana Fernández, SEO consultant [40:08]

If your homepage scores under 40 in the auditor, fix the positioning sentence this week and leave PR for next month. It takes an afternoon, costs nothing and gives you a before-and-after to compare once you measure your brand in SearchBrand.