Yandex Direct keyword operators: quotes, brackets and fixed word forms

If your team knows Google Ads, the first thing you look for in Yandex Direct is the match type setting. Broad, phrase, exact. There isn’t one.

Yandex controls which queries trigger your ad through operators written inside the keyword itself: quotation marks, square brackets and a form-fixing exclamation mark. And the report that looks like it measures match quality hides more than it shows. Let’s take the levers one at a time, and then the more important question of where precision stops helping.

Why the match type report misleads

First, about the illusion of control.

Yandex reports include a category that sounds like word-by-word matching, as if inside it sat queries whose words matched the keyword literally. Check it against the data and the opposite appears.

Break that category down by the distance between the words of the query and the words of the keyword, ignoring stop words and order, and the picture is this: a large majority of impressions fall on rows where the match is partial. Inside “word-by-word” matching sits mostly semantic matching, non-literal matches by meaning.

The label creates the appearance that you are managing precision while you are not even measuring it. Two consequences follow.

The match type report is not a usable measure of how tightly queries hit your keywords. And the familiar argument of “literal versus semantic” loses its meaning: the second is contained inside the first, and the report will not tell you in what proportion.

The same breakdown also buries a popular myth, that the more precisely a keyword matches the query, the cheaper the click. There is no stable relationship of that kind in the data.

Quotes, brackets, fixed forms: what each operator does

Now the tools that actually control what you match. They behave differently, and confusing them is expensive.

Quotation marks are exact phrase matching. Maximum strictness, minimum traffic volume. As the main way to gather volume it is unsuitable: it cuts off too much.

The whole phrase in square brackets fixes word order. It also cuts a great deal: reorderings, words inserted between the given ones. Not recommended as a working default, especially in Russian, where reordering is natural and constant.

Each word in its own square brackets is a separate construction and the most interesting lever. It requires that every specified word be present in the query in any word form, but fixes neither the order nor the absence of extra words between them.

That is the one to understand properly. Because word forms stay free, the nested depth survives and the long tail is not cut. What it does exclude is the substitution of your fixed words by synonyms. And synonym substitution, along with autotargeting’s own substitutions, is exactly what drags foreign traffic into a keyword. This construction closes that door without closing the tail.

A few practical rules:

• Do not create duplicates for different cases of the same word. Brackets already cover every word form. This is where imported Google habits waste the most effort.
• Prepositions can usually stay unbracketed, but check whether the preposition carries meaning. In Russian it often does.
• Fix only the meaningful words: the main category and its significant attribute. Leave geographic add-ons and “hot” words like buy or order free, so you keep collecting traffic through their synonyms.
• The fewer words you fix, the larger the share of traffic arriving through synonyms of the unfixed ones. This is an adjustable width lever, not an on/off switch.

Brands and vendors are a separate story. There you add every product category paired with both the Cyrillic and the Latin spelling of the name, and pin the word forms themselves with the exact-form modifier, an exclamation mark before the word. Without that, brand impressions drown in junk, and for a foreign brand whose name Russians transliterate in two or three ways, this is not optional.

Precision is a lever, not a goal

It is easy to get carried away and fix everything, chasing a perfect match.

The industry lived for years on the belief that maximum precision wins: rigid types, literal coincidence of query, keyword and headline. When that belief was put to the numbers, splitting conversions and cost per lead by degree of match, no stable relationship appeared. Business results are weakly connected to how literally the query hit the keyword.

The lesson carries over to the neighbouring platform. In Google Ads, moving the same keywords to broad match often raises conversions and lowers cost per lead, while manual cross-negation between ad groups tends to hurt: it tries to impose on the model a traffic distribution the model already computes from a vast number of signals.

The point is not to loosen everything blindly. The point is to stop praying to precision for its own sake. Fixing is needed where junk gets in without it. Where there is no junk, extra strictness cuts your own demand and creates work out of nothing. Decide by the account’s actual numbers, not by habit.

What not to spend operators on

Some queries no operator will save, because they should never have been collected.

Those are informational and multimedia formulations: read about, see photos, watch a video. At that stage the person is not buying, and statistically they barely click ads at all.

In Google Ads the mechanism is visible in the auction: expected click-through rate is calculated across the whole industry and carries noticeable weight in quality score. If people generally do not click ads on a query, the platform mostly stops showing paid results on it, so as not to spoil the page for users. The result is that informational keywords deliver near zero traffic and do not repay the collection, the clustering or the maintenance they demand.

Before you configure operators, separate commercial queries from informational ones and simply leave the second group out.

Where else you control what you match

Audience fit is not set by words alone.

Competitor domains, sorted by frequency, work as a separate criterion: brand keywords collected from them go into search campaigns, into the network and into display formats, while the domains themselves become an interest targeting condition. Reach through an allowed-placement list is usually small, so this is a later-stage technique rather than a main one. But as a way to extend the keyword set beyond the obvious, competitor domains give you something operators cannot collect.

What to do with your keyword set this week

1. Stop reading the match type report as a precision meter. It hides semantic matching inside the word-by-word label.
2. Learn per-word brackets as your main lever against synonyms and autotargeting, while keeping the nested depth.
3. Fix only the meaningful words. Leave geography and hot words free for synonyms.
4. Do not worship literal precision. Check on your own data whether match tightness correlates with cost per lead at all.
5. Keep informational queries out of search campaigns. There is almost no paid traffic on them.

Operators are a fine instrument, not a set of spells for maximum precision. They exist to cut one specific kind of junk, synonym substitution, not to strangle your own demand in pursuit of a literal match. Whoever understands that spends brackets sparingly and collects more relevant traffic with less effort.

If you want to see where your keywords catch junk and where you are cutting demand with unnecessary strictness, that is one of the things the free review looks at.

Frequently asked questions

Sources

• Yandex Direct help, keyword operators: yandex.com/support/direct/en/keywords/symbols
• Yandex Direct help, search terms and match reporting: yandex.com/support/direct/en/statistics
• Russian version of this article: Операторы соответствия в Директе