What Is Fill with Gemini in Google Sheets?
Fill with Gemini is an AI feature in Google Sheets that populates cells for you: it reads the context of your table, infers what each column needs, and fills ranges through a drag gesture or a plain-English prompt. According to Google's own 95-participant study, it completes 100-cell data entry tasks 9 times faster than typing the values manually.
If part of your week still disappears into spreadsheet cleanup, retyping categories, standardising dates, copying product details from one format into another, you are doing work this feature was built to remove. The frustrating part is that most eligible users have the button sitting in their toolbar right now and have never clicked it.
Google announced Fill with Gemini on the Workspace Updates blog on 22 April 2026, and it reached most business and consumer AI accounts through May. On 7 July 2026 it expanded to 11 additional languages, which is why many non-English workspaces are only now seeing it appear.
Under the hood, it is an entry point to the AI function in Sheets, the same engine behind the =AI() formula. The difference is that Fill with Gemini needs no formula knowledge at all.
How Do You Use Fill with Gemini?
There are two entry points. Drag-and-drop: if a column has at least one completed cell, drag the fill handle and Sheets offers to complete the column based on the table's context. Prompt-based: select an empty range, click the Fill button that appears, and either accept the automatic fill or type a custom instruction in plain English.
The drag method feels like the classic autofill you already use for dates and number series, except it now understands meaning, not just patterns. Fill one cell with "Complaint" next to an angry customer message, drag down, and Gemini classifies the remaining rows by reading each message.
The prompt method gives you more control. Select the empty range, click Fill, and describe the output you want: "Format these dates as DD MMM YYYY" or "Write a one-line summary of the feedback in column B."
Three practical setup notes:
--- Your table needs headers. Gemini infers intent from column names and neighbouring data. A column labelled "Sentiment" next to "Customer comment" practically fills itself.
--- One good example beats zero. For drag-fill, complete one row by hand first. That single example anchors the format of everything below it.
--- Results arrive as suggestions. You confirm before values are committed, so you can reject a bad fill without damaging the sheet.
What Tasks Does Fill with Gemini Handle Best?
It performs best on four task families: categorising existing text (sentiment, topic, priority), reformatting values (dates, names, phone numbers), extracting structure from messy text (pulling company names or order numbers out of paragraphs), and drafting short text per row (reply suggestions, one-line summaries). Google's launch examples include a marketing team generating suggested responses to customer feedback.
Scenario 1: The survey dump. You export 300 open-text survey responses. Add columns for "Theme" and "Sentiment", fill one example row, drag down twice. What used to be an afternoon of manual tagging becomes a ten-minute review pass.
Scenario 2: The product catalogue. A supplier sends product names in one messy column. Prompt-fill a new column with "Extract the brand name only" and another with "Write a 15-word product description in a neutral tone."
Scenario 3: The CRM cleanup. Names arrive as "CHAN Tai Man", "tai man chan", "Chan, T.M." Prompt: "Rewrite each name as First Last with normal capitalisation." Format chaos ends without a single nested formula.
Notice what these have in common: each row's answer lives inside the row itself. That is the sweet spot. Tasks that need heavy outside knowledge or live web data are where quality gets shakier, treat those fills as drafts to verify, not facts.
How Do You Write Prompts That Fill Cells Correctly?
The reliable pattern is: name the source column, define the output format, and constrain the options. Vague prompts produce inconsistent cells; constrained prompts produce clean columns. "Categorise this feedback" drifts, while "Read column B and label it as exactly one of: Pricing, Product, Service, Other" stays consistent for hundreds of rows.
Weak prompts fail quietly in spreadsheets because you rarely read every cell afterwards. Build your instruction like a specification, not a wish.
Try this prompt (copy-paste into the Fill box):
"Read the customer message in column B. Label this row with exactly one of these categories: Pricing, Product quality, Delivery, Service attitude, Other. Use the exact spelling above. If the message covers two categories, choose the one the customer mentions first."
And a formatting one:
"Convert the date in column C to the format 10 Jul 2026. If the cell does not contain a recognisable date, write CHECK instead of guessing."
That final instruction, giving the model an explicit escape hatch like CHECK, is the single highest-value habit. It converts silent errors into visible flags you can filter for in ten seconds.
Who Gets Fill with Gemini and What Are the Usage Limits?
Fill with Gemini is included in Google Workspace Business Standard and Plus, Enterprise Standard and Plus, the AI Expanded and AI Ultra add-ons, and consumer Google AI Pro and Ultra plans. It is on by default, but disappears entirely if your admin has switched off Workspace smart features. Note one date: promotional higher usage limits run only through 15 July 2026.
After 15 July 2026, per-user usage limits apply to the AI function in Sheets, with higher ceilings for AI Expanded and AI Ultra add-on licences. Google has said updated limits will be documented in the Help Center. Practically: if you plan to batch-process thousands of rows, do the heavy runs now and learn where your plan's ceiling sits.
If you cannot see the feature at all, check three things in order: your plan is on the eligible list, Workspace smart features are enabled in your settings, and your language is covered, remembering the 11-language expansion only started rolling out on 7 July 2026 and can take up to 15 days to reach every account.
Where Does Fill with Gemini Break Down?
Its failure modes are the classic AI ones wearing a spreadsheet costume: plausible-looking values that are wrong, inconsistent labels across a long column, and confident fills on rows where the source data was empty or ambiguous. It is also the wrong tool for auditable numbers, anything a finance reviewer must trace back to a formula or source.
--- Do not use it for figures you must defend. A generated number has no formula trail. For anything auditable, keep using formulas, and use Gemini only to draft or label around them.
--- Long columns drift. Past a few hundred rows, category spelling can wander ("Service" becomes "service issues"). The exact-spelling constraint from the prompt section suppresses most of this; a pivot table afterwards catches the rest in seconds.
--- Ambiguous sources produce confident nonsense. An empty feedback cell can still get a sentiment label. Add "If column B is empty, leave this cell blank" to your prompt.
--- Web-sourced fills need verification. Where Gemini pulls outside knowledge, company details, addresses, spot-check a sample of at least 10 rows before you trust the column.
Try It Now: A Five-Minute Test on Your Own Data
Open any sheet with a column of messy text, feedback, product names, or meeting notes. Add a header called "Category" in the next column, select the first 20 empty cells under it, click Fill, and paste the constrained category prompt from earlier, adjusting the labels to your data. Review the 20 suggestions before accepting.
Twenty rows is deliberate: large enough to expose inconsistency, small enough to check every cell in two minutes. If the labels hold up, run the remaining rows. If they wobble, tighten the option list and try again, you are debugging a specification, not fighting the tool.
Then try one drag-fill: complete a single example cell, grab the fill handle, and watch whether the suggestions match your example's format. Those two tests cover both entry points and tell you exactly how much of your weekly cleanup this feature can absorb.
Conclusion: The Formula Barrier Just Disappeared
Fill with Gemini matters because it removes the skill gate between "people who can write REGEXEXTRACT" and everyone else. The judgement work, defining categories, setting formats, deciding what gets verified, stays with you; the typing goes away. Practitioners who learn to write constrained fill prompts now will quietly reclaim hours before their colleagues notice the button exists.
The best AI features are the ones that meet you inside the tools you already use. We understand AI. We understand you better. With UD by your side, AI doesn't feel cold.
Automate More Than Your Spreadsheets
Cell-filling is the start; whole workflows are the prize. UD helps Hong Kong teams hand routine work to AI staff that run around the clock, and we'll walk you through every step, from tool setup to workflow design and deployment.
