Line chart maker

What a line chart does that a bar chart cannot

A bar chart tells you how big each period was. A line chart tells you what happened between them.

The difference is slope. When values are connected, the eye reads the angle of the connection as rate of change — and it reads it automatically, without counting. That is why a line chart can carry twenty-four months of data comfortably while a bar chart of the same numbers becomes a picket fence, and why "growth is accelerating" is visible in a line chart and invisible in bars.

The cost of that connection is a promise. Joining two points asserts that the values between them moved smoothly from one to the other. When that is not true — when the categories are separate things rather than a continuous progression — the line is lying, and bars are the honest chart.

When to use a line chart

It is the right chart for:

• Anything over time with more than about eight periods — monthly revenue, daily active users, weekly signups
• Comparing two to four series that share a scale — this year against last year, three products against each other
• Showing rate of change where the slope is the finding, not the individual values
• Continuous measures — temperature, price, concentration — where intermediate values genuinely exist

Reach for something else when:

Your situation	Better chart	Why
Fewer than about six time periods	Bar chart	Too few points to show a shape; bars read the values more precisely
Categories with no continuity between them	Bar chart	A line between "North" and "South" asserts a progression that does not exist
Emphasising cumulative volume	Area chart	The fill carries magnitude that a bare line does not
Composition changing over time	Stacked bar	Shows the total and the mix together
More than about five series	Small multiples	Five lines is the practical ceiling before it becomes spaghetti
Two measures on wildly different scales	Two separate charts	Dual axes can be made to show almost any relationship you want

The rule that is the opposite of a bar chart

A line chart does not have to start at zero.

This surprises people who have learned "always start at zero" as a universal rule. It is not universal — it is a rule about bars, because a bar encodes value as length, and a shortened bar shows a false length.

A line encodes change as slope. The line's position on the axis is secondary; what the reader takes away is the shape. Forcing a zero baseline on data that lives between 94% and 97% produces a flat line near the top of an empty chart, which hides the very thing you are plotting.

So the correct rule for line charts is: choose a range that shows the movement, and label the axis clearly so nobody is misled about where it starts. ChartGPT does exactly this — when your values sit in a narrow band, a line chart zooms to that band and the tick labels show where the axis begins. Charts whose values already span a wide range keep the zero baseline, because there is nothing to gain by cutting it.

This applies to line charts only. Bars, areas and stacks encode value as size, so they always start at zero.

The place this goes wrong elsewhere is when a narrow range is used to make a trivial change look dramatic in a chart with a faint or missing axis. The fix is the label, not the baseline.

Three worked examples

1. A single trend

Data:

Month	Signups
Jan	120
Feb	145
Mar	138
Apr	190
May	215
Jun	240
Jul	232
Aug	280

Prompt: Line chart of monthly signups Jan to Aug. The point is that growth picked up from April.

What you get: one line, months in order along the bottom, and an axis running from 100 rather than zero so the March and July dips remain visible. The title states the April inflection rather than naming the axes.

2. Year on year

Data:

Month	FY25	FY26
Jan	120	180
Feb	145	195
Mar	138	220
Apr	190	265

Prompt: Line chart comparing FY25 and FY26 monthly signups

What you get: two lines sharing one axis, each with its own colour and a legend beneath the chart. Keep it to four series or fewer — past that the lines start crossing often enough that the legend stops helping.

3. A metric in a narrow band

Data:

Week	Uptime
1	99.2
2	99.5
3	98.8
4	99.7
5	99.9

Prompt: Line chart of weekly uptime percent

What you get: an axis running 98.5% to 100% rather than 0% to 100%. This is the case the rule above exists for — a zero-based axis would compress five weeks of meaningful variation into a flat line at the top of an empty chart. The tick labels make the range obvious.

Handling awkward time series

Missing periods. A value that is absent is read as zero, and a zero is a real reading — it will pull the line down to the axis and look like a collapse. So do not leave holes: either drop the row entirely, so the period does not appear on the axis at all, or supply the number. This is the one piece of tidying worth doing before you paste.

Irregular intervals. If your points are Jan, Feb, Mar and then Dec, they must not be evenly spaced along the axis. Even spacing would make a ten-month gap look like a one-month gap and distort every slope in the chart.

Too many points. Daily data over three years is a thousand points, and the line becomes a band of noise. Aggregate to weekly or monthly before pasting — and note that three years of daily data is over a thousand rows, past the free plan's 200-row paste limit anyway.

Four things that ruin a line chart

1. Spaghetti. Six or more series and nobody can follow any of them. If you need six, use small multiples — six small charts in a grid, same axes — rather than six lines in one frame.

2. Dual axes. Two measures on two different y-axes can be made to show correlation, divergence or nothing at all purely by choosing the axis ranges. Two stacked charts sharing an x-axis say the same thing without the manipulation. ChartGPT draws one axis.

3. Connecting things that are not continuous. A line joining product categories, survey options or regions implies a progression between them. Those are bars.

4. Smoothing that invents data. A curved spline through sparse points draws values that were never measured, including peaks and troughs between your actual observations. ChartGPT joins points with straight segments, which is the honest version.

Getting your data in

Paste a range from Excel or Sheets, or type the values into a sentence. Keep your headers — the date column header becomes the axis label and the value column headers become the series names.

For line charts specifically, put time in the first column and each series in its own column. Two columns gives you one line, four columns gives three.

The paste limit is 500 rows, on every plan. Line charts are the place it actually binds — three years of daily data is over a thousand rows.

How to make one

1. Paste your data, time in the first column.
2. Say line chart so it does not infer from the data shape.
3. Name your units so the axis and subtitle read correctly.
4. Say what the chart is for — "the point is that growth stalled in Q3" gets you a title that says so.
5. Adjust in words. Keep calendar order. Shorten the title. Use monthly totals instead of daily.
6. Export — SVG for slides and projection, PNG for a quick paste, PDF for print.

Related chart types

• Area chart — a line with the volume beneath it filled in. Use when magnitude matters as much as direction.
• Bar chart — for fewer periods, or for categories rather than time.
• Grouped bar — comparing two or three series across a small set of periods.

Exporting

SVG, PNG and PDF, on every plan including free. SVG for anything projected or resized — line charts suffer badly from raster scaling, because thin strokes are the first thing to blur. See pricing for what the paid plans add.