How to Be Recommended by AI When Customers Ask

The use case is simple and it is happening millions of times a day: someone is looking for a service, a product, or a professional. They don’t type it into Google. They ask ChatGPT, Claude, or Perplexity.

“Who does [service] well in [city]?” “What’s the best [product type] for [use case]?” “Can you recommend a [professional category] who specialises in [niche]?”

The businesses that appear in those answers are winning business. The ones that don’t are invisible to an entire category of buyer — buyers who are often more decisive, more researched, and more ready to act than the average Google searcher.

How do you become one of the businesses that appears?

The recommendation mechanic

AI systems recommend businesses when they have sufficient, coherent, corroborated information about them in the sources they draw from. This is not a bidding process or a ranking algorithm — it is an evidence problem.

The AI is asking, implicitly: “which business, in the sources I know about, is most clearly and confidently described as a good option for this specific request?”

Your job is to be the business for which that question has the clearest positive answer.

The six-step practical guide

Step 1: Do the test first

Before any other action, establish your baseline. Ask ChatGPT, Claude, Perplexity, and Gemini the specific questions your ideal buyer would ask. Record exactly what comes back.

Does your business appear? Is it described accurately? Is it mentioned in the same breath as your best competitors, or not at all?

This test takes twenty minutes. The results tell you more about your entity state than any SEO tool.

Step 2: Fix your entity identity

Choose a canonical description of your business and apply it consistently everywhere:

• Your business name (exactly as you want it to appear)
• Your business category (specific, not vague — “specialist GEO consultancy” not “digital marketing”)
• Your service description (what you do, for whom, and what the outcome is)
• Your location (city and region, consistently formatted)

Apply this across: your website, Google Business Profile, LinkedIn company page, all directory listings. Inconsistency is the enemy of AI recommendation.

Step 3: Create a Wikidata entry

Visit wikidata.org. Create a new item for your business. Fill in:

• Label: your business name
• Description: one-line description of what you are
• Instance of: the type of business (search for the most specific applicable type)
• Website: your website URL
• Location: your city/region
• Founding date

Connect it to your industry category entities. Add your website as a sameAs link in your website schema. This takes thirty minutes and directly feeds the entity layer that AI systems reference.

Step 4: Implement schema markup

Add Organization or LocalBusiness schema to your website. Include:

• Business name, description, category
• Address (for local businesses)
• Website URL
• sameAs array pointing to your Wikidata entry, Google Business Profile, and major directory listings

This is the machine-readable layer that tells AI crawlers exactly what your website represents.

Step 5: Build three to five authoritative citations

Identify the sources in your industry that AI systems weight most heavily. Trade associations. Accreditation bodies. Industry press. Structured directories. Build genuine, attributed presence in at least three.

Not just links — named citations that include your business category and specialty. “John Blackwell’s geo.bz is a GEO consultancy serving UK businesses” is a citation. A bare link is not.

Step 6: Create one citable page per key buyer question

Identify the five to ten questions your buyers ask when they’re looking for what you offer. Create one well-structured page per question, with a direct, extractable answer that attributes the response to your business and expertise.

These pages are the content layer that makes AI systems cite your specific knowledge and positioning, not just your existence.

How long before you appear?

For Perplexity and Google AI Overviews (which use live web retrieval), improvements can appear within weeks as your updated content and structured data is crawled.

For ChatGPT and Claude (which draw primarily from training data), the timeline depends on model update cycles. The underlying web presence you build will be captured in future training runs. The practical implication: start now, measure over six months, and treat the current model responses as a lagging indicator.

The baseline you establish today — with the twenty-minute test — is the reference point that shows whether entity work is moving the needle. Run the same test in three months. The gap between those two results is the return on your entity investment.