How to Get ChatGPT and Perplexity to Recommend Your Brand

Table of Contents

• Quick summary
• Traditional SEO vs. GEO: what changes for your brand
• Why can a brand be left out of the answer?
• What makes an assistant cite one brand and not another?
• 30-day plan to get recommended

  • Days 1-10: audit and technical base
  • Days 11-20: rewrite your key pages
  • Days 21-30: earn mentions off your site
• How do you measure whether it works?
• Frequently asked questions
• Crea Cosas tools to run the plan
• Conclusion
• Sources

More and more buyers ask an assistant "what's the best software for…?" and decide from the answer, without opening ten tabs. If your brand isn't in that answer, it isn't in the comparison. Getting recommended is not a technical trick: it depends on your information being clear and verifiable, and on other trusted sources talking about you.

This guide gives you a 30-day plan, from audit to measurement. If you want the data and limits behind GEO, start with our research on GEO and SEO in the AI era; here we focus on one concrete goal: making your brand the one that gets mentioned.

Quick summary

• The goal changed. Besides ranking, you want to be cited and recommended inside the answer.
• Audit first. 20 real questions across 4 assistants show where you stand today.
• Being in Bing matters. ChatGPT with search retrieves results from Bing's index.
• Answer first, explain after. Direct answers at the top get cited more.
• Off-site beats on-site. Third-party mentions weigh more than any tweak to your own site.
• Nobody guarantees results. Answers vary; measure over time.

Traditional SEO vs. GEO: what changes for your brand

	Traditional SEO	GEO (generative engines)
Goal	Position in the results list and clicks	Be cited and recommended in the answer
Metric	CTR, rankings, traffic	Appearances, share of mentions vs. competitors, cited sources
Content	Pages per keyword	Direct answers, original data, honest comparisons
Authority	Backlinks and on-page optimization	Brand mentions on third-party sites, communities and videos

They are not separate worlds: useful, well-indexed content is still the base of both. That's why most of GEO is well-done SEO plus presence outside your site.

Why can a brand be left out of the answer?

Assistants combine two sources. The first is real-time search: they query the web (ChatGPT uses Bing's index; Perplexity retrieves on its own), extract snippets and use them to answer. The second is what the model learned during training, which you can't control or update at will.

According to a Seer Interactive analysis of 80 million queries, 87% of ChatGPT's citations match results already in Bing's top results. And when a Google AI Overview appears, Pew Research measured that users click a link 8% of the time, versus 15% when there isn't one. If the user doesn't click, being mentioned in the answer is the only presence you get.

What makes an assistant cite one brand and not another?

Nobody outside the platforms knows the exact formula, and any "weights" in percentages you see repeated are reverse engineering. What studies do show:

• Early answer. A Kevin Indig analysis found that 44% of ChatGPT citations come from the first 30% of the page.
• Verifiable evidence. In the Princeton paper that named GEO, adding quotes, statistics and sources worked best; repeating keywords performed worse than doing nothing.
• Outside mentions. Ahrefs, across 75,000 brands, found a 0.664 correlation between web brand mentions and AI Overview visibility, versus 0.218 for backlinks.

One important nuance: models don't reward an "encyclopedia tone" by itself. What helps is that your claim is specific, has a source and matches what others say about you.

30-day plan to get recommended

Days 1-10: audit and technical base

1. Audit with 20 questions. Write 20 questions your ideal customer would ask, for example "what are the best solutions for [problem] in 2026?" or "compare [your brand] with [competitor]". Ask them in ChatGPT, Perplexity, Claude and Gemini. Log in a spreadsheet:

• whether you're mentioned and in what position in the list;
• which competitors appear with you;
• which sources they cite (those are the pages you need to win);
• the tone: positive, neutral or with reservations.

2. Index in Bing. Submit your sitemap to Bing Webmaster Tools. It takes about ten minutes and it's the condition for ChatGPT to find you.

3. Review robots.txt with judgment. AI crawlers fall into training (GPTBot, ClaudeBot), search (OAI-SearchBot, Claude-SearchBot, PerplexityBot) and on-demand user fetches (ChatGPT-User). If you want to be recommended, don't block search or user bots. Blocking training bots is a legitimate data-policy decision, but it doesn't improve your citations.

4. Fix friction. Crawl errors, broken links, redirect loops and slow pages make crawlers give up.

About llms.txt: you can create it, but don't put it ahead of the above. Google has said it doesn't use it, and no major platform documents it as a citation signal.

Days 11-20: rewrite your key pages

Pick the 10 pages that bring the most sales or inquiries and apply this to each:

• Direct answer first. The first two sentences answer the main question.

  Before: "In this article we'll explore the costs of buying business software…" After: "No-code automation software for businesses typically costs between $20 and $50 per user per month on standard plans; enterprise plans reach several thousand."

• Question-style subheadings. Change "Software costs" to "How much does no-code software cost in 2026?".

• Easy-to-extract blocks. Comparison tables, short lists and a summary at the top.

• Original data. A figure from your customers, a real case study, a real price. This is the seeding that works: being the original source of a data point. If you don't have the figure, don't invent it; a fake number destroys the trust you're after.

• Honest comparisons. A "your brand vs. alternatives" page with a table, "best for" and the competitor's real strengths. Assistants tend to cite balanced comparisons over brochures.

• Verifiable authorship. Every article with a real author, bio, experience and links to professional profiles.

• Basic structured data. Article, Organization, Person and Product describe who you are. Don't expect schema alone to make you appear.

If you need to be clear on who you're talking to before rewriting, use our guide to buyer personas with 6 examples. If you sell to businesses, start with the B2B ideal customer profile (ICP).

Days 21-30: earn mentions off your site

This is the phase with the most weight and the one most often neglected. Look to appear, with real value, where your topic is already discussed:

• Reddit and niche forums, answering questions with experience, not self-promotion.
• Third-party comparisons and lists: trade media, directories, Product Hunt, review sites.
• YouTube: the single most cited domain in AI Overviews according to several analyses. Turn your key article into a video with a full transcript.
• Consistent mentions: your name, category and value proposition should say the same thing on every profile.

The formula is simple: "[your brand] + [the specific problem you solve]" in the sources that already showed up in your day-1 audit.

How do you measure whether it works?

Repeat your 20 questions every month and calculate:

• Appearance rate: in how many of the 20 answers you're mentioned.
• Share of mentions: your mentions versus those of your direct competitors.
• Cited sources: which URLs from your domain appear in Perplexity or ChatGPT footnotes.
• Sentiment and position: whether you're recommended first or third, and how you're described.

In your analytics, create a dedicated channel for traffic from chatgpt.com, perplexity.ai and claude.ai, and also watch growth in brand searches and direct traffic: many users see the recommendation and then look you up by name. Answers vary from day to day, so compare trends over several weeks, not a single screenshot. Tools such as Otterly, Peec AI, Ahrefs Brand Radar or Profound automate this, but you can start with a spreadsheet.

Frequently asked questions

Can I pay for ChatGPT to recommend my brand? No. Assistants build answers from what they find on the web and what third parties say about you.

How long until it shows? Technical changes can show up within weeks in assistants with real-time search; outside mentions take months to build.

Do I need llms.txt or FAQ schema? Not as a priority. Neither is a proven citation lever. The most common doubts are covered in Frequently asked questions about GEO.

Crea Cosas tools to run the plan

• Value proposition: organize who you serve and what you offer with the Business Model Canvas template.
• Pricing and margins: check your offer's profitability with the profit margin calculator.
• Technical implementation: for a step-by-step version for creators, follow the practical GEO guide.

Conclusion

Being recommended by an AI doesn't depend on a single tactic, but on three things that reinforce each other: a site assistants can find, pages that answer clearly with original data, and third-party mentions that confirm what you say. Start today with the 20 questions: in an hour you'll know where you stand and which sources you need to win.

Sources

• Original GEO paper (arXiv)
• Google's guide to optimizing for AI
• Pew Research: AI results and click behavior