Where to Find GEO Prompts and How to Work with Them in Practice

When people talk about GEO (Generative Engine Optimization) and generative search results, the conversation quickly comes down to one piece of advice: write articles based on prompts. And this immediately raises a practical question — where to get these prompts. Not as general ideas, but as specific phrasing that generative systems already use to produce answers.

My name is Kirill Yandovskiy, and I have been working in SEO for more than 15 years. When working with content for generative search, one of the key steps is not writing the text itself, but understanding which prompts it actually makes sense to write for in the first place. This is exactly the step that causes the most difficulty.

What is Considered a Prompt in the GEO Context

In the context of GEO, a prompt is not a technical command for a neural network and not a phrase that should be literally inserted into the text. It refers to the way a query is formulated when a user interacts with generative search. This can be a question, a clarification, a comparison, or an attempt to understand a topic — in other words, a query with context and details rather than a short, generalized keyword.

If simplified, a classic SEO query answers the question “what to find,” while a prompt answers the question “what to understand.” The difference is clearly visible in examples:

An SEO query is short and generalized: company formation Dubai.
A prompt is expanded and more specific: how long does it take to open a company in Dubai and what steps should I take.

The meaning may be similar, but the form and level of detail are different.

For generative systems, query volume is not as important as intent clarity. A longer and more precise phrasing gives AI more context and helps it select appropriate sources for an answer. These types of queries are more likely to become the basis for generative search results.

Therefore, when we continue talking about finding prompts, it is important to keep one simple idea in mind: we are not interested in sets of keywords, but in specific context-rich formulations for which generative systems are already building answers.

Where to Find Real GEO Prompts

The main mistake when working with GEO is trying to make up prompts on your own. As a result, you get logical, neat phrases that look good on paper but poorly match how queries are actually formed in generative search. It is much more reliable to rely on sources where such phrases already exist.

Google as a Prompt Source

Google is the simplest and at the same time the most underestimated source of prompts. Its value is not in search volume and not in data from tools, but in how exactly Google shows the way users phrase their queries.

If you look at Google not as a keyword-based search engine, but as an interface for questions, it starts to work as a free prompt database.

Autocomplete in the Search Bar

Autocomplete is the first place where it makes sense to look for prompts. It shows not abstract keywords, but real query suggestions — real ways queries are phrased — that people are already typing.

Important moments:

enter not a single term, but the beginning of a query;
do not rush — watch how the suggestions change as you add words.

For example, if you start typing “how to register a company,” Google may continue with:

how to register a company in dubai
how to register a company as a foreigner
how to register a company step by step
how to register a company online

These are no longer keywords, but ready-made query phrasings that can easily be turned into prompts for an article. Here it is important to capture not only the phrase itself, but also the logic of clarification — for example, country, user status, format, or stages.

“People Also Ask” Block

People Also Ask is one of the most useful sources if the goal is to find prompts specifically for articles.

This block is valuable because:

the questions are already fully phrased;
they are logically connected to each other;
Google shows them not randomly, but as a continuation of the user’s intent.

Examples for the same topic:

Do I need a license to register a company in Dubai?
How long does it take to register a company in Dubai?
Can a foreigner own 100% of a company in Dubai?

Each such question is either a separate prompt or part of a single cluster, based on closely related query wording, for which you can write one article or a large section.

A practical technique: click to expand the questions — Google loads new query phrasings, and in 5–10 minutes you can collect a solid list.

“People Also Search For” Block

The related searches block is often ignored because it looks “SEO-ish.” But if you look at it closely, you can also find clarifying query wording there, not just synonyms.

For example:

company registration dubai cost
company registration dubai requirements
company registration dubai timeline

Prompt Hints in the “People Also Search For” Block

On their own, these are not prompts yet, but they hint at which aspects of the topic people are trying to understand. These terms work well as “anchors” for building more detailed prompt phrasing.

How to Work with Google in Practice

It is important not to try to “pick the best” options right away. The goal at this stage is to collect raw material:

query phrasings,
clarifications,
recurring patterns in how questions are phrased.

Only later, at the filtering stage, does it become clear which of them are suitable for articles and which are better left aside.

Google Search Console as a Source of Prompts

Google Search Console allows you to work not with hypotheses, but with real query phrasing that a website is already appearing for in search results. In the context of GEO, this is useful as a way to understand which prompts are already being formed around the site’s topic and to use them as a foundation for new articles.

Here, we are not looking for a replacement for the semantic core, nor are we concluding that the site already fully covers these queries. On the contrary, GSC helps reveal query wording that:

goes beyond standard SEO keywords;
includes clarifications and context;
can logically be expanded into standalone articles or large sections.

In practice, there are two convenient ways to work with prompts in Google Search Console.

Finding Prompts via the Queries Table

This approach is useful when you want to look at queries not selectively, but as a whole — as a dataset of query phrasing related to the website’s topic. Working with the table makes it easy to quickly separate short SEO keywords from longer, more specific queries that, by their structure, are closer to prompts and suitable for developing new articles.

Step-by-step process:

Open the Performance → Search results → Queries.
Export the list of queries to Excel.
Add a column next to the query column called Query length.
Count the number of characters in each query using the formula: =LEN(A1), where A1 is the cell containing the query. The formula counts spaces, numbers, and symbols.
Drag the formula down for all rows and sort the table by the Query length column in descending order.

At the top of the list, you will see long, more detailed query phrasings — exactly the queries that are structurally closer to prompts than to classic SEO keywords. These can be conveniently used as a basis for new articles or grouped into larger pieces of content.

At the same time, it is important to watch out for potential irrelevant queries. After filtering, additional cleanup may be required.

Finding Prompts by Filtering Question-Based Queries

Filtering by question-based queries is a faster way to find prompts that are already phrased as expanded queries. This approach does not require analyzing the entire list: the filter immediately surfaces query wording that can be treated as a ready foundation for an article or a separate content block.

Step-by-step process:

Open the Performance report → Search results → Queries.
Click + New → select Query.
Choose Custom (regex) as the condition.
Insert the regular expression: ^(how|who|what|where|when|why|which|can|could|do|does|did|is|are|was|were|should|would|will|may|might)\b
Apply the filter and review the resulting list of queries.

From this list, select query phrasings that:

contain context and clarifications;
look like complete, well-formed queries;
are suitable as a foundation for a new article.

If needed, the list can be exported and worked with further in the same way as in the previous approach: grouping similar query phrasings, combining them into a single piece of content, or selecting the most promising prompts.

Reddit as a Source of Prompts

Reddit is convenient for finding prompts because the phrasing here is not constrained by site structure, semantics, or search templates. Users describe their tasks directly, with details and context, without trying to “adapt” to search. This type of language works especially well as a foundation for prompts.

Unlike search engines, Reddit does not smooth phrasing or reduce it to a standardized form. This makes the platform useful for finding natural language that can later be used when planning articles.

Where and How to Find Prompts on Reddit

The primary source of prompts on Reddit is post titles. This is where users try to phrase their query as precisely as possible.

Practical workflow:

Use Reddit’s built-in search with a base term or topic.
Review results specifically by post titles.
Switch sorting to New or Top (for the past month or year).
Write down the phrasing in full, without shortening or editing it.

At this stage, it is important to preserve the original wording — do not rewrite titles for SEO and do not try to “improve” the phrasing.

How to Use Comments

Comments help expand the original prompt and understand which details users consider important. Discussions often include clarifications, alternative ways of phrasing queries, and related questions. These elements can be conveniently used as a basis for article subsections or for logically expanding the main prompt.

How to Select Query Phrasing for Articles

After collecting query phrasings, it makes sense to review which of them:

are repeated across different discussions;
contain clear context and conditions;
can be logically developed into a standalone article.

Reddit is especially useful in cases where you need to go beyond standard search phrasing and understand how people describe their tasks outside of search-driven logic.

Otterly.ai as a Source of Prompts

Otterly is useful because it allows you to obtain prompts not “from your head,” but through specific inputs: a URL, keywords, or a brand. These prompts can then be used immediately as a foundation for briefs for new articles.

Method 1: Finding Prompts via a Competitor’s URL

In Otterly, open the AI Prompt Research section.
Select the mode “I have a specific URL and want to discover which prompts are driving traffic.”
Enter a competitor’s URL (or a specific competitor page), and choose the language and country.
Otterly generates a list of prompts that trigger generative responses involving that site.

You can then put these prompts straight to work: they become the basis for briefs for copywriters (for individual articles or article clusters).

What matters in practice:

it’s better to start not with a competitor’s homepage, but with a specific URL that is already ranking or being cited for the topic;
prompts should be recorded in their original wording, without any improvements.

Method 2: Finding Prompts via SEO Keywords

This mode is useful when you already have a semantic core (or at least a list of base keywords) and need to turn it into prompts for articles.

Open AI Prompt Research.
Select the keyword-based search mode (via SEO keywords).
Enter 1–5 keywords related to a single topic (do not mix different clusters in one query).
Choose the language and country.
You will receive a list of prompts — usually more expanded query phrasings around the core topic.

From there, the logic is the same: prompts → grouping → briefs for copywriters.

Method 3: Finding Prompts via a Brand

This option is useful when you need to understand which prompts are associated with a brand, even if the website does not yet have strong visibility in generative search results.

Open AI Prompt Research.
Select the brand-based search mode.
Enter the brand name, domain (website), and/or business focus (category or niche).
Choose the language and country.

You will receive a list of prompts where the brand appears or where typical queries around it are present.

These prompts work well for:

“brand + topic” articles;
FAQ content;
comparison pages (if they appear in generative results).

How to Turn Discovered Prompts into Articles

After working with Google, Google Search Console, Reddit, and Otterly, you usually end up with a list of prompts. At this stage, it is important not to lose focus and not to slide back into classic SEO, where everything is reduced to keywords and search volume.

The first thing to do is to keep the prompts in their original form. Do not rewrite them for SEO, do not “improve” the phrasing, and do not immediately turn them into headlines. A prompt is not an article title, but an entry point into the article. It defines the logic of how the topic is unfolded, not the final wording of the text.

Next, it makes sense to group prompts by meaning. Typically, several query phrasings describe the same task but with different clarifications. In this case, they do not become separate articles, but rather form the basis of a single piece of content — with logical sections and a coherent sequence of answers.

The next step is to define the format.

One prompt = one article, if the phrasing is narrow and specific.
Several closely related prompts = one article with a clear structure, if the topic requires explanation from multiple angles.

Only after this does it make sense to move on to creating a brief for a copywriter. In a well-prepared brief, prompts:

serve as the foundation of the structure;
define the order in which the topic is covered;
help avoid “blurred” content that lacks a clear answer.

It is important to understand that working with prompts is not a replacement for SEO and not a separate discipline. It is a way to more precisely match the expectations of generative systems and to write articles not “generally about a topic,” but for specific query phrasings that already exist in search and AI-generated results.

In summary, prompt research is not a one-time task and not an intuitive process. It is systematic work with sources, query phrasing, and content structure. And the earlier it becomes part of the workflow, the easier it is for a website to appear in generative results without trying to guess how AI works.

If you have difficulties finding promts, or want to rank on generative search results, or have any questions related to SEO- and GEO-promotion, book a consultation on the website or DM me on Telegram @yandowski.