Search Everywhere: Keyword, Question and Prompt Research for SEO, AEO and GEO
A six-step research workflow: map topics and entities, build a keyword universe, then add the questions and AI prompts your audience actually asks.

This post is Chapter 3 of Search Everywhere: The Practical Guide to SEO, AEO and GEO, the free online edition. Chapter and appendix references link to the matching post.
Every optimisation effort starts with the same question: what are people actually looking for, and how do they phrase it? In classic SEO, the answer was a keyword list. In AEO it is a question list. In GEO it is a prompt list. They are related but not identical, and a modern research process produces all three.
This chapter walks through a research workflow you can repeat for any site, using webocreation.com/’s core topic, OpenCart, as the worked example.
Search intent comes first
Before looking at any numbers, understand intent: the goal behind a query. Most frameworks use four categories.
- Informational: the searcher wants to learn. “What is OpenCart”, “how to install OpenCart 4”, “opencart vs woocommerce”.
- Navigational: the searcher wants a specific site or page. “opencart admin login”, “webocreation opencart”.
- Commercial investigation: the searcher is comparing options before buying. “best opencart hosting”, “opencart developer cost”, “top opencart themes 2026”.
- Transactional: the searcher is ready to act. “hire opencart developer”, “buy opencart theme”, “download opencart turnstile extension”.
The intent decides the format of the page that can win. Google tells you which format it prefers: just search the query and look at what ranks. If the top ten results are all step-by-step tutorials, a service page will not rank there however well you optimise it. If they are all comparison lists, a single product review will struggle.
AI assistants add a twist. Because prompts are long and conversational, a single prompt often mixes intents: “I’m building a jewellery store, should I use OpenCart or WooCommerce, and roughly what would a developer charge?” contains informational, commercial and transactional elements. Pages that serve the whole journey (explain, compare, then offer a next step) are well placed for these prompts.
Step 1: Map your topics and entities
Start by listing the topics you have a genuine right to cover. For a site to be trusted by both search engines and AI systems, it needs a clear topical identity: an area where it is obviously an expert.
Case study. webocreation.com/’s strongest topical claim is OpenCart. The founder has written four books on it, the site has hundreds of OpenCart tutorials and more than 80 free extensions. Secondary topics with real depth are HubSpot, Marketo and Pardot templates, Webflow, and Cloudflare security for online stores. Weaker, scattered topics include blockchain, NFTs, health-sector policy notes and internship reports.
Write your topics as entities where possible. An entity is a distinct thing that search engines and knowledge graphs recognise: “OpenCart” (software), “Cloudflare Turnstile” (product), “Core Web Vitals” (concept), “Rupak Nepali” (person). Entities connect to other entities. When your content consistently connects your brand to the right entities, both search engines and language models learn the association.
A simple topic map for the case study might look like this:
OpenCart (core)
├── OpenCart 4 (versions, upgrade from 3, compatibility)
├── Extensions (install, develop, free downloads, OCMOD)
├── Themes (install, customise, Twig templates)
├── Store setup (products, categories, payments, shipping, tax)
├── Performance and security (Cloudflare, caching, backups, captcha)
├── SEO for OpenCart (SEO URLs, schema, sitemaps, Search Console)
└── Development (MVC-L, events, API, database)
Marketing automation (secondary)
├── HubSpot (HubL, email templates, landing pages)
├── Marketo
└── Pardot
Web platforms (secondary)
├── Webflow
└── WordPress
Step 2: Build the keyword universe
Now gather keywords for each branch. Use several sources, because each has blind spots.
Your own Search Console data. This is the most valuable source because it shows queries you already appear for. Export the Performance report filtered to the last 16 months, with queries and pages. Queries with high impressions but low clicks and positions between 5 and 20 are your quickest wins.
Semrush Keyword Magic Tool. Enter a seed such as “opencart” and you get thousands of related keywords with search volume, keyword difficulty (KD), intent labels, CPC and SERP features. Filter by “Questions” to pull question-format queries, which feed your AEO list. Chapter 18 covers the tool in depth.
Semrush Organic Research on competitors. Enter a competitor’s domain and see the keywords they rank for. For OpenCart topics, competitors include the official OpenCart documentation and forum, extension marketplaces, hosting companies’ tutorial sections, and other development agencies.
Google’s own suggestions. Autocomplete, People Also Ask, “People also search for” and related searches at the bottom of the page all show real phrasing. Tools such as AlsoAsked and AnswerThePublic collect these at scale.
Google Trends. Use it to compare terms (“opencart” vs “woocommerce” vs “shopify”) and to see seasonality and rising queries.
Communities. Reddit, Stack Overflow, the OpenCart forum, GitHub issues, and YouTube comments show the problems people actually have, in their own words. These are gold for AEO and GEO because they mirror how people write prompts.
Your own customers. Support emails, contact form messages and sales call notes. Every question a customer asks is a question someone else is typing into a search box or an AI assistant.
Step 3: Classify and prioritise
Put everything into a spreadsheet with these columns:
| Column | Purpose |
|---|---|
| Keyword | The query |
| Topic cluster | Which branch of your topic map it belongs to |
| Intent | Informational, navigational, commercial, transactional |
| Volume | Monthly searches (from Semrush or Keyword Planner) |
| Difficulty | KD score |
| SERP features | Featured snippet, PAA, AI Overview, video, etc. |
| Current position | From Search Console or Position Tracking |
| Target URL | Existing page, or “new” |
| Priority | Your score |
A simple priority score multiplies business value (1 to 3: does this keyword lead to customers?) by opportunity (1 to 3: can we realistically rank or improve?) and divides by effort (1 to 3: how much work?). Do not over-engineer it. The purpose is to make decisions, not to produce a beautiful spreadsheet.
Illustrative example. Part of a prioritised list for webocreation.com/. Volumes, difficulty and positions are made up for illustration.
Keyword Intent Vol. KD SERP features Pos. Priority opencart 4 captcha Info 170 18 PAA 7 High opencart turnstile Info/Trans 90 12 none 3 High upgrade opencart 3 to 4 Info 480 31 FS, PAA, Video, AIO 11 High opencart developer Trans 720 44 Local, Ads 28 Medium opencart vs woocommerce Commercial 1,300 52 FS, AIO 34 Medium nft profitable methods Info 2,400 61 AIO – Ignore (off-topic)
The last row is deliberate. High volume is not a reason to target a keyword that does not fit your topical identity. Off-topic traffic rarely converts, and off-topic content can dilute how search engines and AI systems understand what your site is about.
Step 4: Build the question list (AEO)
AEO targets questions. From your keyword universe, pull every query that is phrased as a question or that clearly implies one. Then expand it.
Sources of questions:
- Semrush Keyword Magic Tool, “Questions” filter.
- People Also Ask boxes. Each time you expand a question, Google loads more. Tools such as AlsoAsked map the full tree.
- Search Console queries containing “how”, “what”, “why”, “can”, “does”, “is”, “vs”, “best”.
- Forums and Q&A sites.
Group questions by the type of answer they need, because each type has a winning format:
| Question type | Example | Best answer format |
|---|---|---|
| Definition | What is OCMOD in OpenCart? | One- or two-sentence definition, then detail |
| How-to | How do I install an extension in OpenCart 4? | Numbered steps |
| Comparison | Is Turnstile better than reCAPTCHA? | Short verdict, then a comparison table |
| List | What are the best free OpenCart themes? | Bulleted or numbered list with one line each |
| Yes/no | Does OpenCart 4 support PHP 8.3? | “Yes” or “No” first, then the conditions |
| Troubleshooting | Why is my OpenCart admin showing a blank page? | Most likely cause first, then a checklist |
| Cost/time | How much does an OpenCart developer cost? | A range, then what drives it |
Chapter 21 turns this table into writing patterns.
Step 5: Build the prompt list (GEO)
Prompts differ from keywords in three ways: they are longer, they carry context about the person, and they often ask for a recommendation. You cannot get prompt volumes from any tool in the way you get keyword volumes, because AI companies do not publish them. So you build a representative list instead.
How to write a prompt list:
- Start with your personas. For webocreation.com/: a store owner who is not technical, a freelance developer, an agency, a marketer working with HubSpot or Marketo, a student.
- For each persona, write the problems they would bring to an AI assistant. Use the language of your customer emails and forum posts.
- Write each as a realistic prompt, including context. Not “opencart captcha”, but “I run a small OpenCart 4 store and I’m getting fake customer registrations every day. What’s the best way to stop them without annoying real customers?”
- Cover the funnel. Include learning prompts (“explain how OpenCart extensions work”), comparison prompts (“OpenCart or WooCommerce for 5,000 products?”), and recommendation prompts (“who are good OpenCart developers in the US?”, “best books to learn OpenCart 4 development”).
- Include brand prompts. “What is webocreation.com/?”, “Who is Rupak Nepali?”, “Is webocreation a good OpenCart developer?” These test what models know about you.
Aim for 30 to 100 prompts for a small site and several hundred for a large brand. This becomes the fixed test set you run regularly to measure AI visibility (Chapter 24).
Illustrative example. A starter prompt set for webocreation.com/.
- What’s the easiest way to stop spam registrations on OpenCart 4?
- How do I upgrade from OpenCart 3.0.3.8 to OpenCart 4 without losing orders?
- Which books should I read to learn OpenCart 4 extension development?
- Recommend free OpenCart 4 extensions for social sharing and related products.
- Is OpenCart still a good choice for a small store in 2026, compared with WooCommerce and Shopify?
- How do I set up Cloudflare WAF rules for an online store?
- How do I build a coded HubSpot email template with custom modules?
- Who is Rupak Nepali?
- Can you recommend an OpenCart developer who also does SEO?
- How do I fix “Crawled – currently not indexed” on my eCommerce product pages?
Step 6: Map keywords, questions and prompts to pages
The final step turns research into a content plan. For each topic cluster, decide:
- Which existing page is the best target for each keyword group. One primary page per intent; avoid having several pages compete for the same query (keyword cannibalisation).
- Which questions that page should answer, usually as H2 or H3 sections.
- Which prompts the page should satisfy, which tells you what context, comparisons and recommendations to include.
- Which new pages are needed to fill gaps.
- Which existing pages should be merged, redirected or removed because they compete with each other or do not fit the site.
Tip: the cannibalisation check. In Search Console’s Performance report, filter by a query (for example “opencart captcha”) and switch to the Pages tab. If several of your URLs share impressions for the same query and none ranks well, they may be competing. Consolidate them into one stronger page and redirect the others.
Tools summary
| Need | Free option | Paid option |
|---|---|---|
| Your own query data | Google Search Console, Bing Webmaster Tools | (none better) |
| Keyword volumes and difficulty | Google Keyword Planner (ranges) | Semrush, Ahrefs |
| Question research | People Also Ask, AnswerThePublic (limited), AlsoAsked (limited) | Semrush Questions filter, AlsoAsked Pro |
| Competitor keywords | Manual SERP review | Semrush Organic Research, Keyword Gap |
| Trends and seasonality | Google Trends | Semrush trends data |
| Prompt research | Your support inbox, forums, Reddit | Semrush AI visibility tools, dedicated AI tracking tools |
Key takeaways
- Start with intent and topical identity, not search volume.
- Build three lists from the same research: keywords for SEO, questions for AEO, prompts for GEO.
- Group questions by answer type, because each type has a winning format.
- Prompt lists have no volume data. Build them from personas and real customer language, and keep them fixed so you can measure change.
- Map everything to pages, and fix cannibalisation and off-topic content before creating more.
Previous: How Search Engines and AI Answer Engines Work
Next: Crawlability and Indexability: robots.txt, Sitemaps, Canonicals and Redirects
All chapters: Search Everywhere: The Practical Guide to SEO, AEO and GEO






