Search intent: what it is and how to read it from the SERP

The four jobs hiding behind one search phrase

The cells carry no axis labels because there are no axes: each cell is read through its own example.

Informational

Wants to understand. "what is SEO". The answer is a guide.
Navigational

Knows the destination. "search console login". The answer is that site itself.
Commercial

Is comparing. "best hosting in Iran". The answer is a comparison.
Transactional

Has decided. "buy WordPress hosting". The answer is a product page.

one phrase

The borders between these cells are soft and one phrase can sit in two of them; what decides is the dominant format on the results page, not this grid.

Last checked: 6 September 2026 Facts and tool names in this lesson are re-checked against their sources on this date.

What does search intent actually mean?

Search intent is the job a user wants done behind a phrase, not the topic written in the phrase. "what is web design" and "web design price" share one topic and ask for two completely different things: the first wants an explanation, the second wants a number.

Four buckets do the work in practice. Informational, when someone wants to understand something: "what is SEO". Navigational, when they already know the destination and only want the door: "search console login". Commercial investigation, when they have not decided yet and are comparing: "best hosting in Iran". Transactional, when the decision is made and they want to finish the job: "buy WordPress hosting".

Those four are a working model, not a law of Google. Google's own quality rater guidelines cut it differently, with categories named know, do, website and visit-in-person. Both are models and neither is meant to be complete. What matters is the part you can see: for each intent Google lifts a different kind of page to the top, and the kind of page is visible.

Why no tool reports intent for a Persian keyword

English-language tools carry a column called intent, with a label sitting next to every keyword. For Persian there is no such thing, and the gap is not in our tooling; the data itself does not exist.

We measured this with live API calls rather than reading documentation. DataForSEO's Labs product lists ninety four locations and not one of them offers fa among its available languages. Its search intent endpoint works from a language code and answers for English, but fa and the names Persian and Farsi all come back as an invalid field. Keyword difficulty sits in exactly the same hole.

There is a temptation here worth naming out loud: the advertising competition value is always available, it moves between low, medium and high, and it looks like difficulty. That value counts bidders, not how hard a ranking is. Printing it under a heading that says difficulty means printing a wrong number that looks confident, which is worse than an empty column.

The practical result for Persian: you have search volume, you have cost per click, and intent and difficulty are yours to read. The next section is that reading.

An English keyword and a Persian keyword take different routes

Does intent data exist for this keyword's language?

Yes, English for example

Take the label from the tool

The search intent endpoint answers from a language code
Keyword difficulty lives on the same route
Still open the results page for your head terms

No, Persian

Read it yourself off the results page

Count the dominant format in the first ten results
Write down the blocks on the page
Leave the difficulty column empty, not filled with an ads number

The Persian route is slower, but what it gives you beats a machine label, because it comes from today's actual results page.

Read the intent off the results page

The method is simple. Search the phrase in a private window, set to the country and language your audience actually uses, then look at the shape of the page instead of the rankings.

What you are looking for is the format that dominates the first ten results. If most of them are guides and explainers, the intent is informational and a sales page does not fit there. If "best of" lists and comparisons are on top, the user is still choosing. If product pages and shopping ads fill the space, the decision is made and the job is to finish it. And if a brand's own site sits at the top with sitelinks, the phrase is navigational and competing for it wastes time.

Two more signals are worth a look: the related questions block tells you what people ask next, and a map block means Google treats the phrase as local. The anatomy of the results page, and what each block does and does not let you control, is opened up in the anatomy of a Google results page.

A warning about the words themselves: "what is", "how to", "price" and "buy" are good hints but not verdicts. A phrase like "wordpress hosting" carries no verb at all and only the results page reveals its intent. When your guess and the page disagree, the page wins.

Which signals to lean on and which ones mislead

Reading intent off the results pagesignals

Lean on these

The format repeated across the first ten results
Whether shopping ads and product pages are there at all
A map block, which means Google treats the phrase as local
The phrasings that show up in the related questions block

Do not lean on these alone

Words like buy and price sitting inside the phrase
The advertising competition value standing in for difficulty
A label a language model produced without seeing the page
Your own guess about what the customer wants

The second column is not wrong, it is insufficient: each of these is sometimes right, which is exactly what makes it dangerous.

What happens when your page misses the intent

When the page type misses the intent, it usually does not rank at all. The worse case is that it does rank: the visitor lands, works out in a few seconds that this is not what they wanted, and leaves. That visit is not a sale and it says nothing good about your page either.

The common Iranian example is a service page optimised for "what is X". Someone typing "what is" is not a customer yet, and a price page is no use to them. The right answer is two pages doing two jobs: one explains, one sells, with a natural link from the first to the second.

Our position here is blunt: do not convert an educational page that works into a sales page to squeeze more revenue out of it. That traffic arrived for exactly the reason the page explains something. Put one clear path to the service page and leave the rest alone. And if you already have two pages answering the same intent, your problem is not intent, it is keyword cannibalization.

One phrase can carry more than one intent

Some phrases do not carry a single intent. The results page for "wordpress hosting" holds both guides and product pages, because Google is not sure which one the user wants and shows both so they can choose.

Do not build a hybrid page for this. A page that teaches and sells at once usually satisfies neither intent completely. Pick the larger share, answer that one fully, and send the other intent to the right page with a link.

One last thing that saves a lot of time: intent does not stay still. Close to a buying season the results page for some phrases turns more commercial, and months later it turns back. So before writing any page, open the results page for its phrase once. It takes ten minutes and prevents a month of writing the wrong thing.

The fast path, with AI

Labelling intent across a two hundred keyword list is mechanical work, and a language model genuinely frees up time here. The condition is that the model has not seen the results page and cannot see it, so its role is linguistic sorting, not judgement. A fast, cheap model of the Flash class is enough for that, and our current pick is listed in the <a class="text-link" href="/en/ai/">AI section</a>.

Export the keyword list with its search volumes and leave it as it is; the model must not add a keyword or a number.
Label five to eight keywords by hand yourself, a few from each of the four intents. Those examples are the expensive part, and without them the output is not trustworthy for Persian.
Fill the recipe below with the list and the examples, then run it.
Open in Google only the rows the model returned with low confidence, plus the head of each group. You verify those few, not all two hundred.

Copy-ready recipe

Role: Persian SEO specialist. Your job is labelling only. You may not add a keyword, drop a keyword, or produce any number.

Four allowed labels: informational, navigational, commercial, transactional.

My hand labelled examples (these are your standard, not your general knowledge):
{e.g. what is SEO = informational | best hosting in Iran = commercial | buy WordPress hosting = transactional | search console login = navigational}

The keyword list, one per line: keyword | search volume
{list}

Return one line per row, with exactly these columns:
keyword | volume | label | confidence high or low | the page type this intent wants

Rules:
1. If the phrase has no verb and could carry two intents, mark confidence low. Do not guess.
2. If the phrase is the name of a specific brand or service, label it navigational.
3. "price" alone is not a transactional signal; if the phrase is still comparing, label it commercial.
4. At the end, repeat only the low confidence rows in a separate list so I can check them in Google myself.

Before you trust the output: The output of this recipe is a default that cuts the manual work from two hundred rows to ten, not a finding. The model has not seen the results page and will not spot mixed intent unless your examples taught it to. Persian adds one more error layer: commercial Persian phrases often carry no verb, and without examples the model reads them as informational. Open the head of each group and every low confidence row in Google yourself.

AI in this kind of work

A simple division of labour works here: the language model sorts the list, you confirm the intent. The reverse, handing the intent decision to the model, is where this breaks, because the model is not connected to Google and has not seen the results page you are looking at.

Tools that actually help

Google Search Console The strongest intent evidence for your own pages: the phrases you actually get shown for. If a sales page collects impressions on "what is" phrasings, the intent is written right there.
Gemini A good fit for bulk labelling of a list, on its fast cheap model. Google's own page says the Gemini web app runs in over 230 countries and territories, and Iran is not on that list.
Claude Better on the borderline rows, because it will explain itself and the explanation can be argued with. Iran is not on Anthropic's supported-countries list and there is no official payment route from Iran.
RGB The access and payment layer for Iran, kept separate from the tools themselves.

Where it backfires

The main risk here is not a visible mistake, it is unearned confidence. The model returns a label for every row, including rows whose honest answer was "the results page is mixed", and those labels later become a content calendar that nobody re-checks. Persian raises the risk, because commercial Persian phrases usually carry no verb and without examples the model reads them as informational. And the worst case: ask a model which sites rank for a keyword and you will get an answer, and that answer is fabricated. Google's own guidance is clear on the frame: the question is not how content was produced, it is whether it is useful to the person who searched.

Sources: Google Search Central: creating helpful, reliable, people-first content Google Search Central: Google Search and AI-generated content DataForSEO Labs: search intent endpoint (language based, no Persian)

Where this advice stops

Intent is not fixed and this lesson is a snapshot. The layout of a results page differs between mobile and desktop and between cities, and it changes without notice, so every reading has a shelf life. More importantly, the results page tells you what Google rewards today, not what your business should sell; a phrase whose intent does not match your work is not your phrase, even when you read it correctly.

From our own work

We have implemented this data gap in our own tooling and it can be inspected. In the demand tables on this site, rgb_dm_metrics() in themes/rgb/inc/demand.php drops the difficulty and intent columns entirely whenever a Persian row is in the table, rather than printing them empty. The next decision was harder: we had the advertising competition value and could have dropped it into the difficulty column, and we did not. A live measurement on 2026-08-27 showed the Labs product carries ninety four locations and none of them supports fa, so any number we put there would have been made up. A missing column is more honest than a wrong one.

Real follow-up questions

How many types of search intent are there?

The common model has four: informational, navigational, commercial and transactional. Google's own quality rater guidelines split it differently, with different names. Both are models and there is no fixed count; what decides is the results page for that particular phrase.

Where do I get intent for a Persian keyword?

From no tool, because the data does not exist for Persian. Open the results page for the phrase in a private window and count the dominant format in the first ten results. For long lists, hand the first pass to a language model with hand-written examples and check only the doubtful rows yourself.

What search intent is

Informational

Navigational

Commercial

Transactional