Table of Contents
TL;DR
- Search intent beats search volume for a small site. The keyword whose intent matches a page you can write honestly earns its keep. The keyword whose intent does not match bounces.
- Four intents cover almost everything: informational, navigational, commercial, transactional. Most keyword lists mix all four — the intent label is what separates a useful page from a wasted one.
- Commercial-intent prompts trigger ChatGPT web search 53.5% of the time. Informational ones trigger it only 18.7% of the time. Intent is the strongest single predictor of whether the chat surfaces will go look for your page at all.
- AI chats label intent well and fabricate search volumes badly. Use a chat for the intent sort and the clustering. Get volume from Search Console or a real keyword tool.
- A small-site keyword list belongs on one screen. Twenty to thirty keywords, each tagged with intent, each mapped to one page, each volume confirmed at a real tool.
Profound’s January 2026 analysis of ChatGPT behavior put a number on something most small-site owners already feel. Commercial-intent prompts trigger a ChatGPT web search 53.5% of the time. Informational-intent prompts trigger one only 18.7% of the time.
That single gap explains why two pages on the same site, with the same word count and the same author, can pull wildly different traffic. The intent of the keyword decides whether the chat surface even goes looking for a page to cite.
A small site cannot beat a national chain on keyword volume. The chain has more pages, more backlinks, more brand mentions in the Knowledge Graph. The small site can still win on intent match. Intent match means picking the keyword whose underlying customer goal lines up with a page you can write honestly, today, with what you already know.
This piece is the short version of how that filter works. The four intents you need to recognize. The thirty-second read for any keyword. The mechanical fix for the one big mistake every AI chat keeps making with keyword data.
Why does search intent matter more than keyword volume for a small site?
A keyword tool will tell you the volume. It will not tell you whether the visitor is the visitor you want.
Volume on its own does nothing. A page that ranks for "what is SEO" gets a thousand visitors a month and one inquiry a year. A page that ranks for "small business SEO consultant Austin" gets thirty visitors a month and three inquiries a week.
Both pages might show the same green bar in a keyword tool. The intent behind the search is the difference.
A small site has limited writing time. Twenty hours a month, maybe forty if the owner is also the writer. The right question is not "which keyword has the highest volume." It is "which keyword brings the visitor I can actually help."
The intent of a keyword answers that. Informational means the visitor wants to learn. Commercial means the visitor is comparing.
Transactional means the visitor is ready. Each of those calls for a different page, written in a different voice, ranked alongside different competitors. A small site that picks the right intent for what it actually offers wins more often than a small site that picks the higher-volume keyword.
What are the four search intents you actually need to recognize?
Informational, navigational, commercial, and transactional. Four labels cover almost every keyword you will see in a tool.
Informational keywords ask a question or seek an explanation. "What is search intent." "How does keyword research work." The visitor wants to learn. The right page is a guide that teaches. The conversion target is a newsletter signup or a return visit, not a sale.
Navigational keywords name a specific site or brand the visitor already wants to reach. "Slobodan Dekanic blog." "WP Rocket pricing page." The visitor knows where they want to go. The right page is the one with that exact name.
There is rarely a competition. The brand owns the keyword.
Commercial keywords compare options before a purchase. "Best WordPress hosting under $10." "WP Rocket vs Perfmatters." The visitor is evaluating. The right page is a comparison or a review, written honestly, with the tradeoffs called out. These are the keywords most likely to trigger an AI chat search and most likely to convert.
Transactional keywords mean the visitor wants to act. "Buy WP Rocket." "Hire SEO consultant Austin." The right page is a product page, a sales page, or a contact form. The visitor is ready — the page should not be in the way.
How do you read intent from a keyword in under thirty seconds?
Type the keyword into Google. Look at what is on page one.
The shape of the first three results tells you the intent the engine has already decided on. Long guides and educational posts at the top mean Google has classified the keyword as informational. Product pages and pricing pages at the top mean transactional.
Comparison posts, reviews, and "best of" roundups mean commercial. A specific brand homepage means navigational.
The People Also Ask box adds a second layer. Read the four questions Google shows. They reveal the sub-intents the engine thinks the visitor might also have.
If the questions all assume the visitor is buying, the intent is commercial or transactional. If they all ask for definitions or explanations, the intent is informational.
The exercise takes thirty seconds per keyword. Twenty keywords cost you ten minutes. The output is a working intent label for each one, derived from what the engine already shows.
You did not need a tool to do it. You needed to look at what was already on the screen.
What does an AI chat add that a keyword tool cannot?
A keyword tool gives you volume, difficulty, and a CPC estimate. It rarely gives you intent. The classification work has to happen somewhere.
That is where an AI chat earns its keep. Paste fifty keywords into the chat. Ask for the intent of each in one word. The chat returns a tagged list in seconds.
The clear cases will be right. "Buy WP Rocket" will come back transactional. "What is search intent" will come back informational. "Best WordPress hosting" will come back commercial.
The chat will miss on the ambiguous middle. Commercial-investigational queries trip the classification, especially when the keyword could be a comparison or a how-to depending on who is searching. The chat will guess. The site article on classifying search intent inside an AI chat workflow walks through the prompt that fixes most of these edge cases.
The clustering work is where the AI chat saves real time. Paste fifty keywords. Ask for three to five thematic clusters.
The output is a rough topic map you can refine in fifteen minutes. A keyword tool would take an afternoon of spreadsheet work to produce the same map.
Why do AI chats hallucinate keyword volumes and how do you catch it?
Ask any AI chat for the monthly search volume of a keyword. It will give you a number. The number will look plausible. It is also fabricated.
The AI chat has no live access to Google’s volume data, to Ahrefs, to SEMrush, or to any third-party keyword index. It pattern-matches a number that looks like the kind of number a human SEO would expect. The number is wrong, often by an order of magnitude, and the chat has no way to know it is wrong.
The fix is mechanical. Use the chat for intent labels, clustering, and rephrasing customer questions into search-style queries. Get every volume number from a real source.
Search Console queries are free and reflect actual visits to your own site. Ahrefs and SEMrush both have free trials and paid plans for the SMB tier. Google Keyword Planner is free with a Google Ads account, even if you never run a campaign.
The pattern to follow is simple. The AI chat does the qualitative work. The keyword tool does the quantitative work. Mixing the two is what produces a keyword list that ranks nowhere because the volumes were fiction.
How do you turn a customer question into a keyword the engines actually search for?
Customers do not type the way SEO tools list keywords. They ask in full sentences.
Open your support inbox. Pull the last fifty messages. Read the first sentence of each. The exact wording your customers used is the wording the engines now have to match against.
Ahrefs documented this in its 2026 Brand Radar methodology. People Also Ask boxes are derived from real user search behavior. The questions in those boxes are already in natural-language form. The chat surfaces use these question shapes when they decompose a user prompt into sub-queries.
The translation step is short. Take the customer question. Strip the niceties. Keep the noun, the verb, and the qualifier.
A long question like "which WordPress hosting works best for a small Etsy shop on a budget" becomes a tight search phrase. Something like "best WordPress hosting small Etsy shop budget." That is the keyword. The engines now have to look for a page that answers exactly that question.
A small site has an advantage no keyword tool can replicate here. The owner already knows what customers ask. The translation from question to keyword is fifteen minutes of reading old emails. The site article on no-budget keyword research methods that work for tiny sites covers the full mechanical sweep, including the People Also Ask harvest.
What does a small-site keyword list actually look like once intent is filtered in?
Twenty to thirty keywords. Each one tagged with intent. Each one mapped to one page on your site, existing or planned.
Volume confirmed at a real tool. The list fits on one screen.
That is it. Most keyword lists I see for small sites have three hundred entries. The owner copies them out of Ahrefs or SEMrush and feels productive.
None of them get written. The list is too big to face.
A list of twenty-five keywords can be written in a year by a one-person business. The owner publishes one piece every two weeks. Twenty-five real pieces, each tied to a real customer question, each with confirmed volume, each with the intent matched to what the page can deliver. The site grows because the writing is honest and the targeting is tight.
The columns to add to the list are minimal. Keyword. Intent label. Volume from real source.
Existing URL or "new." That is four columns. Anything more is decoration. The point is to make the list shippable, not to make the spreadsheet look professional.
Other questions worth answering
Why do classifiers stumble on local buyer phrases?
Because local buying phrases mix two motivations at once. Someone typing ‘plumber Austin’ might want a service today or might shop around first.
ChatGPT and other AI assistants will guess at the intent of these phrases. By 2026, the failure pattern is well documented — commercial-vs-informational confusion is sharpest on local queries. Pair every local phrase with a Search Console check before you trust any auto-label.
How do embeddings cluster phrases differently from overlap analysis?
Embeddings group phrases by meaning, not by shared words. Two queries with zero word-match can sit in the same group when they describe the same buyer goal.
Overlap analysis splits them. Cosine similarity between page-level embeddings reveals boundaries, isolated pages, and over-concentration that simple word-overlap misses. Owner-operators running a 2026 workflow find the practical payoff on category pages where ChatGPT and Perplexity both pull synonyms into one retrieval.
When should you refresh the buyer-motivation table you built last year?
After every quarter that reshapes your offer or product line. The buyer goal behind a phrase drifts when products change, seasons shift, or competitors redo the page-one results.
Re-run the SERP check on the keywords that earned visits last quarter. Pair the SERP read with Search Console click data and update the table cell. In 2026, this twice-a-year discipline beats letting the table go stale for an entire year.
How does sub-query expansion change what phrases you target?
Sub-query expansion forces you to think in comparison and follow-up shapes, not head terms. Modern engines break a single prompt into many sub-queries before they retrieve.
Build phrase variants the engines actually search. Add comparison phrasings, attribute combinations, and follow-up questions next to each head term. A 2026 workflow that maps to fan-out beats a 2018-era export on what ChatGPT and Google end up citing.
How should you turn search intent into your next three articles?
Pick the three keywords from your list whose intent matches a page you can write honestly today.
Write each one as a short H1 question that mirrors the customer wording. A real H1 like "How should I pick a WordPress host for a small Etsy shop on a budget?" beats a corporate H1 like "Best WordPress Hosting Solutions for E-commerce." The first is what a customer typed. The second is what an SEO consultant guessed.
Answer the question in fifty words at the top of each piece. Put the supporting detail underneath. The chat surfaces lift the answer block first.
The supporting detail is what convinces the reader to stay. Three articles published in a month is more than most small sites publish in a year. The intent filter is what keeps each one honest.
If you want a second pair of eyes on the intent filter before you write, you can contact me here. Send your draft list of twenty-five keywords. I will mark the three I would write first, with the intent reasoning. No pitch.