Breaking down a prompt for a bilingual Uzbek and Russian content plan
A bilingual content plan is not one plan translated twice. Here is a prompt that accounts for the demand gap, and where it goes wrong.
The full prompt text
"You are a content strategist. Build a three-month content plan for a company blog in Uzbekistan. Inputs: niche, services, target audience, a list of Russian queries with volumes, a list of Uzbek Latin queries with volumes, and publishing frequency per week. Requirements: for each article specify the language, a headline phrased as a search query, the primary query, the format (guide, breakdown, comparison, checklist, question answer), the funnel stage (awareness, consideration, decision), and which other article in the plan it links to. Do not duplicate a topic across both languages when the Uzbek query has its own phrasing with a different meaning — make two separate articles instead. If data for a language is insufficient, say so rather than inventing volumes. Return a table with a publication-week column."
The prompt is long but every line earns its place. Without the funnel stage, plans skew towards easy informational topics that gather traffic and produce no enquiries.
Why you cannot just translate the plan
A mirrored plan is the most common failure of bilingual blogs. People search differently: what is phrased as a price question in Russian may be searched as a how-it-is-done question in Uzbek, and those are different articles, not translations.
Some topics exist in only one language. There are queries almost never typed in Russian, and the reverse. The instruction not to duplicate when meaning differs is what surfaces that gap.
The opposite case also exists: a topic where both versions are needed almost verbatim. That is one article with two language versions and correct cross-linking, not two articles.
Breaking down the key requirements
Requiring headlines phrased as search queries saves the plan from attractive but irrelevant titles. Left alone, the model gravitates towards things like 'Secrets of successful promotion', which nobody searches for.
The format column sets structure and length: a checklist and a comparison are written differently, and without it an editor reinvents the structure every time.
The 'links to' column turns a list of topics into a connected structure. Internal linking designed at the planning stage works far better than links added afterwards.
Realistic publishing frequency
Ask for a plan at the pace you will actually sustain. Two articles a week in two languages is eight pieces a month including native proofreading — already a lot for a small team.
One good article a week for a year beats ten in the first month followed by silence. Search rewards consistency, and an abandoned blog whose last post is six months old lowers trust in the whole site.
Build in slack: one week in four with no new article, spent updating an older one instead. Updates often buy traffic growth more cheaply than new posts.
Typical generation failures
First, the model invents volumes. If you supply no numbers, it will produce plausible ones. The instruction to flag missing data helps, but you still verify in Wordstat.
Second, Uzbek headlines come out as calques of the Russian, recognisable by structure. Ask the model to phrase each headline from the Uzbek query rather than translating, and have a native speaker check.
Third, everything lands at the awareness stage. State the proportion explicitly: roughly half awareness, a third consideration, the rest decision.
What to do with the finished plan
Run every headline through Wordstat with the regional filter and through search autocomplete. Remove or rephrase topics with zero volume in both languages.
Move the plan into a working sheet with status columns: in progress, in review, published, updated. A content plan without statuses stops reflecting reality within a month.
Revisit the plan quarterly using Search Console data: which articles gather impressions, which do not, and which queries arrived unexpectedly. That last group is the best source of topics for the next cycle.