Proven method to position yourself in AIs (13 steps)

Sebastián Álvarez

Autor

Co-founder of CreceRank | +15 years of experience in SEO and Marketing

🌐

Sebastián Cardozo

Co-Autor

Co-founder of CreceRank | +15 years of experience in SEO and Marketing

🌐

If you run a business, it has already happened to you: someone asked ChatGPT for suppliers in your category and your brand did not appear.

You know about SEO. You have the site positioned. But the game has changed fields and no one told you where the new one is.

This article is the map. The complete method we use for AIs to mention and recommend a brand, in 13 steps, with concrete tools and without magic solutions.

A warning before we start: if someone sells you «appear in ChatGPT in 30 days guaranteed», end the meeting. This works, but it works like the gym: with consistency and method, not with secret formulas.

What we are going to do (and why in this order)

Before going into the step-by-step, understand the logic of the method.

When someone asks an AI, «What brand do you recommend?», the model does three things: it searches for sources, evaluates which brands are mentioned with authority, and drafts a recommendation.

The method tackles each of those three things, in order:

• Phase 1, Diagnosis (steps 1 to 5): we find out where you stand today. What do AIs say about your brand, what keywords have real demand, what prompts are we going to monitor, and whether they speak well or poorly of you. Without this initial snapshot, everything else is guessing.
• Fase 2, Fundamentos (pasos 6 y 7): hacemos que tu sitio sea legible para la IA y que cada página responda de verdad. Es la parte técnica y de contenido propio.
• Phase 3, Authority (steps 8 to 11): we build the signals that make the AI choose you: mentions in the sources it already cites, a solid brand both on and off the site, and unique content that no one else can publish.
• Phase 4, Distribute and Measure (steps 12 and 13): we multiply each piece across all platforms and measure week by week if the work moves the needle.

The order is not decorative: each phase feeds into the next.

Optimizing content without diagnosis is like throwing darts with your eyes closed, and building authority on a site that bots cannot read is building on sand.

The method is followed in order, phase by phase. Once installed, measurement never turns off.

Now, step by step.

Step 1: Define objectives, expectations, and timeline (before touching anything)

The number 1 mistake in GEO is starting to produce content without defining what «winning» means.

Define three things in writing:

• Objective: Do you want to appear in the answers of your category? Correct what AI says wrong about you? Gain share of voice over a specific competitor?
• Metric: percentage of relevant prompts where your brand appears, average position, and sentiment. Numbers, not feelings.
• Realistic timeline: a brand with history and authority sees movement in weeks or a few months. A new brand, with little digital footprint, needs 12 to 18 months of sustained work.

Why commit to something so long? For the value of the prize: the visitor coming from AI is worth 4.4 times more in conversion than that from traditional search.

They arrive determined, because AI has already done the pre-sale for them.

Phase 1: Diagnosis (steps 1 to 5)

Step 2: Audit the real situation, not just the GEO

Before optimizing anything, take the complete picture in three layers:

• GEO: what do ChatGPT, Gemini, and Perplexity say about your brand today? Do they mention you? Do they say anything incorrect? Run your 10 most important business questions and note everything down. You can start with our AI visibility checker.
• SEO: What positions are you in on Google? It matters because AIs pull their sources from the search engine index: for AI Overviews, 76% of citations come from the top 10.
• Market: Who is the leader in your category in AI responses? What did they do to get there? That analysis gives you your competitor’s plan.

The analogy is simple: it’s the market study you would do before opening a store. No one rents the place before checking who is already selling in that block.

Step 3: Look for volume in SEO, because there are no reliable data in GEO

Here’s an uncomfortable truth about the ecosystem: there is no reliable source for the «prompt volume» in AIs. Any tool that sells you those exact numbers is estimating.

The black belt solution: use SEO search volume as a proxy.

If 500 people per month google «best accounting software for SMEs», a growing part of that same demand is already asking ChatGPT.

The concrete process, with Ahrefs, Semrush or similar:

• List the categories and problems your company solves.
• Get the search volume and difficulty of each associated keyword.
• Cross them in a matrix and start where there is real volume with low difficulty.

Step 4: Turn those keywords into prompts

Keywords are the raw material; prompts are what you will monitor every week.

The difference matters. No one writes «accounting software SMEs» to ChatGPT. They write: «What accounting software do you recommend for an SME with 10 employees in Mexico that invoices electronically?»

Build your set of 15 to 30 prompts by combining:

• Commercial intent: «What is the best X for Y?», «What X do you recommend?»
• Scenario and constraints: budget, team size, country, industry. Real questions bring context.
• Comparison: «X or Z? Which is better for…?»
• Brand: «What do you think of [tu marca]?», «Is [tu marca] reliable?»

Step 5: Measure sentiment: what you are facing

It’s not enough to know if you are mentioned. It matters how you are mentioned.

If the AI says «it’s an option, although users report poor support,» you have a different (and more urgent) problem than not appearing.

Think of it as an infinite focus group: the AI summarizes everything the internet thinks about your brand.

If the sentiment is negative, the source is usually from reviews, Reddit, or old notes. There is your list of fires to put out before investing in visibility.

Phase 2: Fundamentals (steps 6 and 7)

Step 6: Make your site visible to AI (technical first)

If AI crawlers can’t read your site, everything else is just decorating a window with the curtain down.

The 6 points of the checklist:

• robots.txt: check that GPTBot, PerplexityBot, ClaudeBot, and Google-Extended are not blocked. It happens more often than you think, especially after migrations.
• Sitemap.xml: complete, with no 404s, containing your important URLs.
• Clean URLs: /prices better than /page?id=8237. The URL structure is part of the message.
• HTML Content: many AI bots do not execute JavaScript. What only exists after rendering in the browser does not exist for them. Verify this by looking at the source code (Ctrl+U), not the inspector.
• Server-side Schema: Organization, Article, FAQPage, and Product as applicable.
• Calibrated anti-bots: an aggressive firewall or captcha may be serving a verification page to AI crawlers. Check the server logs.

And the famous llms.txt? Create it, it takes 10 minutes. But don’t expect anything from it: today almost no bot reads it. It’s at the end of the checklist for a reason.

Step 7: Optimize each page to answer, not to fill

The golden rule that comes from all studies: the answer first, the context later.

Winning structure for each important page:

• Direct answer at the top: what it is, for whom, how much it costs. In the first lines, not in paragraph 14.
• Evidence in the middle: data, comparisons, cases.
• FAQs at the end: with FAQPage schema, answering the real sub-questions.
• Formats that add up: short explanatory videos, summaries, comparison tables, surveys, and charts. Each format is another way for you to be found and cited.

A technical warning that almost nobody gives: comparison tables must be HTML, never images. An image table is invisible to AI. The same applies to prices and features in graphic banners.

And the right volume: length almost does not correlate with citations.

A page that answers 5 specific sub-questions in 800 words beats a 3,000-word manifesto.

The complete detail is in our 7-step model.

Phase 3: Authority (steps 8 to 11)

Step 8: Discover what sources AI uses (and go find them)

Every time an AI recommends brands, it is reading from somewhere. Your job is to find out where.

The process: run your business prompts and note every cited source. In about 20 prompts, you will see the pattern: the same listings, the same media, the same Reddit threads always appear.

That list is your public relations plan. Not «getting mentions in general»: getting mentioned in those 10 specific sources.

We did this exercise at scale in our Experiment 1 on Spanish sources: the «best of» listings dominate recommendations, and UGC (Reddit, reviews) is becoming increasingly important.

Step 9: Build brand relevance within your site

AI needs to understand who you are before recommending you. And most sites make it impossible: a «About us» section of three lines written in 2019.

What to build:

• «About us» with substance: history, mission, team with real names and photos, company data (foundation, size, clients).
• Directors’ pages: background, specialty, media appearances, linked LinkedIn. This feeds the E-E-A-T that AIs inherited from Google.
• Verifiable success cases: with named client, problem, solution, and numbers. «We increased X’s sales by 40% in 6 months» is quotable; «we help our clients grow» is fluff.
• Visible authors on the blog: each article signed by a person with credentials, not by «admin».

Step 10: Build brand relevance outside your site

If you are the only one talking about your brand, for the AI, you do not exist. You need others to say your name.

The tactics, in order of effort:

• Listings in your category: those you found in step 8. Write to the authors, offer access to your product, exclusive data, or a demo.
• Industry interviews and podcasts: each appearance is a co-mention of your brand alongside trustworthy people.
• Guest posts with data: not «5 productivity tips», but articles with your numbers and your experience, signed by your directors.
• Press releases when there is real news: launches, studies, milestones. Fluff press is read by no one, not even the AI.

Paso 11: Crea contenido único y difícil de replicar

Here is the real competitive advantage, the one that cannot be copied with a subscription to an AI tool:

AI does not cite companies. It cites experts.

When ChatGPT needs a piece of data, it looks for the original source: the study, the survey, the experiment. Whoever publishes the original data becomes the reference for the entire category.

What to produce (choose one per quarter, done seriously):

• Own experiments: try something, document the method, and publish the results with graphs.
• Surveys of your market: 100 responses from your industry are data that no one else has.
• Anonymized internal data: trends you see in your customers, benchmarks of your vertical.
• Category research: analyze 50 competitors, 1,000 prices, 200 AI responses. Publish the ranking.

We eat our own food: the numbered experiments of this blog exist for this reason, and they are our most cited and linked contents.

Phase 4: Distribute and measure (steps 12 and 13)

Step 12: Multiply each study across all platforms

A study published only on your blog is a concert in an empty hall.

The logic is press conference: the finding is one, the channels are all. From each study come:

• The full blog article (the citable source).
• A long video on YouTube explaining the findings (remember: YouTube is one of the most cited by AIs).
• Reels and Shorts with the most impactful data.
• A LinkedIn carousel with the graphs.
• A downloadable PDF for press and newsletters.
• A press release to industry media.
• A thread on X and posts in the communities where your audience lives.

Each format is a different door through which an AI (or a journalist, or a client) can find your data and cite you.

Step 13: Measure week by week, like a scientist

AI responses change between runs. Asking once and drawing conclusions is like measuring TV audience by looking at just one house.

The minimum protocol:

• Weekly cadence: enough to see trends, without going crazy with daily noise.
• Multiple runs per prompt: the same question, multiple times, in each LLM. What you’re looking for is frequency of occurrence, not a single snapshot.
• The 4 LLMs separately: ChatGPT, Gemini, Perplexity and AI Overviews behave differently. Appearing in one does not imply appearing in the others.
• Three data points per run: Am I appearing and in what position? What do they say about me (sentiment)? What sources and competitors do they mention?

Doing it manually is possible in the first few weeks. After that, it becomes a full-time job, and that’s where automation comes in.

In CreceRank you set up your prompts once and the platform runs them every day across the 4 LLMs, with position, sentiment, cited sources, and comparison against competitors.

And don’t forget the other half of the measurement: the traffic that the AIs are already sending you. We leave you the guide to measure it in GA4 with the regex ready to copy.

→ Start 14 days free, no credit card required

The complete method, summarized

Conclusion

Positioning yourself in AIs is not magic or luck.

It’s a method that resembles more the SEO of 15 years ago than any trend: those who build authority now, while the majority hesitates, will be the ones that AI recommends for years.

If you take three things from this article, let them be these:

• Diagnosis before tactics: without knowing where you stand (mentions, sentiment, sources), any effort is blind.
• AI cites experts, not logos: your own studies, cases with numbers, and visible people are your defensive moat.
• What is not measured, does not exist: weekly, several runs, each LLM separately.

The first step takes 10 minutes: run your business prompts and see what today’s AIs say about your brand.

In CreceRank you can do it for free and have your baseline this week.

FAQs

Sources

• SaaS AI Search Optimization (Semrush): the 8 signals, the times per task, and the conversion value data of 4.4×.
• Answer Engine Optimization (Ahrefs): the AEO framework, authority factors, and the timelines of 12 to 18 months for new brands.