Shadow Queries Explained: The Hidden Searches AI Uses When Recommending Local Businesses

# Shadow Queries Explained: The Hidden Searches AI Uses When Recommending Local Businesses

The first time I asked ChatGPT “best plumber in Marbella open Sunday”, I assumed it was reading my query, searching the web, and returning the result.

Then I asked it to explain what it did. It told me it ran six separate searches before giving me the answer: one to find plumbers in Marbella, another to check which ones were open on Sundays, a third to verify they were still in business, a fourth to check recent reviews, and two more I can’t even remember now.

That was the moment I understood shadow queries.

## 1. What a Shadow Query Actually Is

A shadow query is a question an AI asks itself, silently, in the background, before it answers your visible question.

### The one-sentence definition

Shadow queries are the hidden sub-searches an AI runs to verify, filter, and rank the results it shows you.

### Why they’re called “shadow” — invisible to every tool

You don’t see them. Google Search Console doesn’t log them. Your analytics doesn’t track them. The AI doesn’t show you its working.

The only way to know they happened is to ask the AI directly: “What did you check before answering that?”

### The difference between a shadow query and a regular AI prompt

A regular prompt is what the user types: “best dentist in Marbella open Saturday”.

Shadow queries are the background searches the AI runs to answer that prompt:
– Is this business open on Saturdays?
– Is it actually in Marbella or nearby?
– Are the reviews recent?
– Do people say it’s good for families, or for cosmetic work, or for emergencies?

The user doesn’t ask those questions. The AI does.

## 2. What It Looks Like in Practice

Let me show you a real example from my testing.

### User asks ChatGPT “best plumber in Marbella open Sunday”

I typed that exact query into ChatGPT on 9 April 2026. It gave me three business names, explained why each one was recommended, mentioned that all three had “recent positive reviews”, and noted that one specialised in “emergency callouts”.

Then I asked: “What searches did you run to find those businesses?”

### Behind the scenes, ChatGPT silently runs 6-10 sub-queries to build the answer

ChatGPT told me it had:
1. Searched for plumbers in Marbella
2. Filtered for businesses with Sunday opening hours
3. Checked recent reviews to confirm they were active
4. Looked for mentions of “emergency” or “callout” services
5. Verified NAP (name, address, phone) consistency
6. Checked whether any were mentioned on local forums or press

Those six searches happened in the background. The user never saw them. I only knew because I asked.

### Example sub-queries for that single plumber question

– “plumber Marbella Sunday hours”
– “emergency plumber Marbella reviews”
– “[business name] still operating 2026”
– “[business name] service area Marbella”
– “plumber Marbella contact details”
– “best plumber Marbella forum mentions”

### The ones you never saw — opening hours, emergency callout, Spanish reviews, local directory presence

And here’s the part that matters: if your business doesn’t answer those hidden questions somewhere on the web (GBP, your site, reviews, directories), the AI filters you out before the user ever sees you.

You’re not competing for the visible query. You’re competing for the shadow queries underneath it.

## 3. Why Shadow Queries Break Traditional SEO Thinking

Everything you’ve learned about keyword research and Search Console doesn’t apply here.

### Keyword research tools don’t see them

Ahrefs, Semrush, and Google Keyword Planner show you what people type into Google. They don’t show you the silent questions ChatGPT asks in the background.

The keyword “best plumber Marbella” has search volume. The shadow query “is this plumber still operating 2026” has zero — because nobody searches for it. The AI does.

### Search Console doesn’t log them

Google Search Console shows the queries that brought traffic to your site from Google Search. It doesn’t log the queries ChatGPT ran when deciding whether to recommend you.

You could be invisible to ChatGPT and your Search Console would never show it.

### Analytics doesn’t show them

Your website analytics logs visitors and referrers. It doesn’t log the AI searches that happened before the visitor even saw your name.

### You can only reverse-engineer them by asking the AI

The only way to discover your shadow queries is to:
1. Ask ChatGPT (or Perplexity, or Gemini) your main local query
2. Ask it “what did you check before answering that?”
3. Take notes

Then answer those questions on your website.

## 4. The 10 Shadow Queries Every Local Costa del Sol Business Faces

I’ve tested this across 30+ Costa del Sol businesses in different categories (restaurants, lawyers, dentists, estate agents, plumbers, tour operators). The shadow queries are always the same 10.

### “Is the business still open?”

The AI checks for recent activity: fresh reviews, recent GBP posts, mentions dated within the past 6 months. If your most recent review is 14 months old, it assumes you might be closed.

### “What is the primary service they offer?”

It reads your GBP category, your website headline, and the first paragraph of your About page. If those three sources say different things, it gets confused and skips you.

### “What town or neighbourhood do they actually operate in?”

“Serving Marbella and surrounding areas” is too vague. The AI wants to know: Marbella Old Town? Puerto Banús? Nueva Andalucía? San Pedro? If your site doesn’t name neighbourhoods, it can’t match you to neighbourhood-specific queries.

### “What do recent customers say about them?”

It reads review text, not star ratings. A 4.2-star business with quotable recent reviews beats a 4.8-star business whose last review is generic or old.

### “Are they licensed or certified for this work?”

For regulated professions (dentists, lawyers, estate agents, financial advisors), the AI looks for licence numbers, professional body memberships, or credentials. If you’re licensed but don’t mention it, the AI treats you as less credible than a competitor who does.

### “What hours do they operate?”

It checks GBP hours. If they’re not set, or if they’re inconsistent with what your website says, you get filtered out of “open now” and “open Saturday” queries.

### “Do they respond quickly?”

It looks at GBP Q&A response times, review reply rates, and mentions of “quick response” or “replied within an hour” in reviews. Responsiveness is a trust signal.

### “Are they recommended by any local community or news site?”

Press mentions, local blog features, Reddit threads, and forum posts all count. If the only place your name appears is your own site and GBP, the AI sees no external validation.

### “What languages do they speak?”

On the Costa del Sol this matters. English-speaking tourists search in English. Spanish locals search in Spanish. The AI looks for explicit mentions: “We speak English”, “Hablamos español”, or reviews that say “the staff spoke perfect English”.

If your site doesn’t mention languages, you’re invisible to one half of the market.

### “Are they more than 5km from the searcher?”

For “near me” queries, the AI uses GBP location data and mentions of landmarks or neighbourhoods. If your address is set but you never mention the neighbourhoods you serve, you get filtered out of hyper-local searches.

## 5. How to Answer Shadow Queries on Your Own Website

Most Costa del Sol business websites are marketing fluff: “We are passionate about delivering excellence.” That’s noise to an AI. It wants facts.

### A dedicated “About Us” section that reads like an entity summary

Write a single paragraph that answers the top 5 shadow queries in plain language:

“We’re a family-run dental practice in Marbella, operating since 2018. We’re located on Avenida Ricardo Soriano, 400 metres from the Corte Inglés. Open Monday to Friday 09:00–18:00, Saturdays 09:00–13:00. We speak English, Spanish, and German. Dr. Elena Martínez is registered with the Colegio de Dentistas de Málaga (licence MA-4521). We typically respond to appointment requests within 2 hours during business days.”

That’s 60 words. It answers: still operating, primary service, exact location, hours, languages, credentials, responsiveness.

ChatGPT reads that and knows you match.

### FAQ schema that mirrors the 10 shadow queries

Add an FAQ section to your homepage or service page and wrap it in FAQ schema (JSON-LD). The AI reads schema directly.

Example for a Fuengirola estate agent:

“`json
{
“@context”: “https://schema.org”,
“@type”: “FAQPage”,
“mainEntity”: [{
“@type”: “Question”,
“name”: “What areas of the Costa del Sol do you cover?”,
“acceptedAnswer”: {
“@type”: “Answer”,
“text”: “We specialise in property sales and rentals in Fuengirola, Mijas Costa, Benalmádena, and Calahonda. We also cover surrounding areas within 15km of Fuengirola town centre.”
}
}, {
“@type”: “Question”,
“name”: “Are you licensed?”,
“acceptedAnswer”: {
“@type”: “Answer”,
“text”: “Yes. We are registered with the Junta de Andalucía as a licensed estate agency (registration number API-MA-1234). All our agents hold API certification.”
}
}, {
“@type”: “Question”,
“name”: “Do you speak English?”,
“acceptedAnswer”: {
“@type”: “Answer”,
“text”: “Yes. We speak English, Spanish, French, and German. All consultations and contracts are available in your preferred language.”
}
}]
}
“`

ChatGPT reads that and can answer user queries like “licensed estate agent Fuengirola English-speaking” without ever visiting your site.

### Operating hours and service area named clearly

Don’t bury your hours in the footer. Put them in your About section, in your GBP listing, and in schema.

Don’t say “serving the Costa del Sol”. Name the towns: “We serve Marbella, San Pedro de Alcántara, Puerto Banús, Nueva Andalucía, Estepona, and Benahavís.”

### Languages spoken named explicitly

Add one sentence: “We speak English, Spanish, and French.” Or “Hablamos español e inglés.”

That’s all it takes. ChatGPT reads it and matches you to bilingual queries.

### Credentials and licences on view

If you’re licensed, certified, or a member of a professional body, say so. Include the registration number if you have one.

“Dr. García is registered with the Colegio de Médicos de Málaga (licence MA-8821).”

“We are members of the API (Agentes de la Propiedad Inmobiliaria) and the AIPP (Association of International Property Professionals).”

The AI treats this as a trust signal. Competitors who don’t mention credentials get filtered out.

For more on structured data, read my [schema markup guide](/structured-data-schema-markup/).

## 6. How to Test Your Own Shadow Query Visibility (10-minute workflow)

You don’t need to hire anyone. You can test this yourself right now.

### Step 1: Ask ChatGPT your main local query

Open ChatGPT and type your primary query:
– “best [your service] in [your town]”
– “best [your service] [your town] open [day]”
– “[specific thing you do] near [landmark]”

Example: “best physiotherapist Fuengirola open Saturday”

Check: Are you mentioned by name?

### Step 2: Ask it “what did you check before answering that?”

Type: “What information did you use to decide which businesses to recommend?”

ChatGPT will tell you. Sometimes in detail, sometimes in summary. Either way, you’ll get a list of 5-10 factors it considered.

### Step 3: For each sub-question, check whether your site answers it

Go through the list and check:
– Does my GBP listing answer this?
– Does my website answer this?
– Do my reviews mention this?
– Is this answered anywhere on the web that mentions my business?

If the answer is no, that’s a gap.

### Step 4: Fix the gaps

Add the missing information to your About page, your FAQ, your GBP description, or your schema markup.

Then test again in 2 weeks. ChatGPT pulls fresh data from the web every time you ask, so changes show up quickly.

For a full audit, I offer a [free AI visibility report](/free-ai-visibility-report/) that tests your business across ChatGPT, Perplexity, and Gemini and tells you exactly what’s missing.

## 7. Shadow Queries vs GEO vs AEO

The terminology around AI search optimisation is messy. Here’s how it all fits together.

### The terminology landscape in plain English

– **AEO** (Answer Engine Optimisation) = optimising for any search engine that gives direct answers instead of links (Google featured snippets, ChatGPT, Alexa, Siri)
– **GEO** (Generative Engine Optimisation) = optimising specifically for AI engines that generate conversational answers (ChatGPT, Gemini, Perplexity, Claude)
– **Shadow queries** = the hidden sub-searches those AI engines run before generating the answer

### GEO (Generative Engine Optimisation)

GEO is the umbrella term for all the tactics you use to get recommended by AI search engines. That includes: citation density, review optimisation, FAQ schema, entity clarity, fresh mentions, and answering shadow queries.

Read my [GEO guide](/generative-engine-optimisation/) for the full framework.

### AEO (Answer Engine Optimisation)

AEO is broader — it includes Google’s featured snippets and People Also Ask boxes, voice assistants, and AI chat engines. If you’re optimising for “how Google answers questions”, you’re doing AEO.

My [AEO guide](/answer-engine-optimisation/) covers the overlap.

### Shadow queries are the plumbing underneath both

GEO and AEO are the strategies. Shadow queries are the mechanism.

When ChatGPT decides whether to recommend you, it runs shadow queries. When Google decides whether to feature you in a snippet, it evaluates similar signals (freshness, entity clarity, structured data).

Understanding shadow queries gives you the “why” behind both.

## 8. The Costa del Sol Advantage

This is an unfair advantage right now, and it won’t last forever.

### Most local competitors haven’t heard of any of this

I’ve audited 40+ Costa del Sol businesses in the past 6 months. Three of them had FAQ schema. Two mentioned their licence number on their homepage. One named the neighbourhoods they served.

The rest? Generic marketing copy, thin GBP listings, no schema, no clear service area, no mention of languages spoken.

They’re optimised for 2015 SEO. They’re invisible to AI search in 2026.

### The first 50 businesses to answer their shadow queries own the AI recommendations for years

AI search is cumulative. Once ChatGPT starts recommending you, it reinforces the pattern: more people ask about you, more reviews mention you, more Reddit threads name you, ChatGPT sees those signals and recommends you more.

The businesses that get in early compound. The ones that wait get filtered out by the early movers’ citation density.

### A specific example: how a Fuengirola dentist went from “not recommended” to “top 3” in 6 weeks

I worked with a dentist in Fuengirola earlier this year. Before we started, ChatGPT didn’t mention them at all for “best dentist Fuengirola” or “dentist Fuengirola English-speaking”.

We did four things:
1. Rewrote the About page to answer the 10 shadow queries (90 words, plain factual language)
2. Added FAQ schema with 8 questions
3. Updated GBP description to match
4. Asked 5 recent patients to leave reviews that mentioned “speaks English” and named the street

Six weeks later, ChatGPT recommended them as one of three dentists for “dentist Fuengirola English-speaking”. Perplexity recommended them first.

No backlinks. No blog content. Just answering the shadow queries ChatGPT was already asking.

## 9. Frequently Asked Questions

### Do shadow queries apply to Google too?

Yes, in a different form. Google doesn’t run conversational shadow queries like ChatGPT, but it evaluates similar signals when deciding what to show in the Map Pack, featured snippets, and People Also Ask boxes.

The tactics overlap: entity clarity, fresh reviews, schema markup, named service areas, hours, languages, credentials.

### How do I find the shadow queries for my business?

Ask ChatGPT your main local query, then ask “what did you check before answering that?” Take notes. Repeat for 3-5 variations of your query (different days, different neighbourhoods, different customer types).

You’ll get a list of 8-12 questions. Those are your shadow queries.

### Will this stop mattering when AI changes?

The specific shadow queries might shift, but the underlying principle won’t. Every AI needs to:
1. Discover which businesses exist
2. Filter them by relevance
3. Rank them by trust

As long as those three steps exist, answering the AI’s background questions will matter.

### Can I pay to rank in shadow queries?

No. Shadow queries are answered by your content, your reviews, your GBP data, and third-party mentions. There’s no ad slot you can buy.

The only shortcut is to answer the questions faster and more clearly than your competitors.

## 10. What to Do This Week

If you want to get visible to AI search, start here:

1. **Test your visibility.** Ask ChatGPT “best [your service] in [your town]” and see if you’re mentioned. If not, ask “what did you check before answering that?” and take notes.

2. **Audit your About page.** Does it answer the 10 shadow queries in plain language? If not, rewrite it. 60-90 words is enough.

3. **Add FAQ schema.** Pick 5-8 of the most common questions people ask about your business and answer them with schema markup. Use the examples in section 5.

4. **Check your GBP.** Is your description factual and entity-clear? Are your hours set? Do your posts show recent activity? Fix the gaps.

5. **Run the test again in 2 weeks.** ChatGPT pulls fresh data. You’ll see movement faster than you would with traditional SEO.

If you want me to run the audit for you, I offer a [free AI visibility report](/free-ai-visibility-report/) for Costa del Sol businesses. No signup, no pitch — just the audit and a breakdown of what to fix first.

—

Shadow queries are invisible, but the businesses that answer them are not.

Most of your competitors on the Costa del Sol have never heard of this. The ones who act now own the AI recommendations for the next 3-5 years.

You don’t need a big budget or a long timeline. You just need to answer the questions ChatGPT is already asking.