Plan Winter Indoor Growing with AI: Choose Plants from Window Light and Room Temperature

When outdoor growing ends in autumn, it is tempting to move pots to a windowsill and keep basil or lettuce going through winter. Yet leafy greens on a winter windowsill tend to produce small leaves, thin stretched stems and pale color. The room looks bright enough, so why do they struggle? One reason is that the brightness your eyes perceive and the amount of light a plant can use for photosynthesis are different quantities. If you do not pin down that difference in numbers before asking an AI for advice, the conditions you hand the AI will already be off.

This article contains affiliate links. As an Amazon Associate, we earn from qualifying purchases.

This article walks through a planning process for winter indoor growing. It covers how to measure window light with a smartphone and where that breaks down, how lux, PPFD and DLI relate, and how to think about winter room temperature and humidity. You then put those conditions into a table, give it to a chat AI such as Claude, ChatGPT or Gemini, and ask for candidate plants and placements. Finally, you check the answer against publications from US university Extension services. The AI proposes candidates; public horticulture references decide. That division of labor is the core of this piece.

Note that we did not measure or grow anything on a real windowsill for this article. Every number comes from a cited source or from a worked example derived from one, and the assumptions behind each example are stated.

Why plants weaken on a winter windowsill: it looks bright

One of the most common problems on a winter windowsill is simply too little light. The Nebraska Extension houseplant guide (NebGuide G2205) lists the symptoms of insufficient light as small leaves, spindly stems and lighter than normal color. Leggy basil on a winter sill fits that description exactly.

The trouble is that low light is easy to miss. Because perceived brightness says nothing about usable light, a winter room at midday still feels bright. What matters to the plant is the total amount of photosynthetically useful light that reaches the leaves over a whole day, and that does not track what your eyes report.

Light requirements also differ widely between plants. Iowa State University Extension (ISU below) groups indoor plants by daily light integral (DLI) as follows.

The key point is that lettuce and basil sit in the high light group, and ISU also lists leafy crops in the very high group. Even comparing the lower bounds, foliage plants start at 3 while leafy greens start at 12, four times as much light. A windowsill where a pothos thrives will not necessarily grow salad. Winter planning starts from that gap.

Purdue Extension (HO-238-W) groups greenhouse crops as 3–6, 6–12, 12–18 and above 18, so the medium and high boundaries differ slightly from ISU. This article uses the ISU groups because they target indoor plants.

Lux, PPFD and DLI: what each unit measures

Three units come up when talking about light. Knowing what each one measures keeps you from getting lost when you give conditions to an AI and when you check its answer.

Illuminance (lux) measures human perception

The lux value shown by a light meter or smartphone app is brightness weighted to the sensitivity of the human eye. Purdue HO-238-W describes foot-candles (the same kind of unit as lux) as a photometric unit based on visible light as perceived by people, mostly reflecting green light, and therefore not suited to describing plant photosynthesis. ISU likewise says illuminance is the amount of light the human eye perceives, not the amount of light in the wavelengths useful to plants.

Because lux is weighted to the eye, the same lux reading can correspond to different amounts of usable light depending on the spectrum of the source. The Cornell CEA hydroponic lettuce handbook also points out that light meters are built to mimic human vision, so they drift from the light available for photosynthesis.

PAR and PPFD count the light plants can use

Plants photosynthesize with light between 400 and 700 nm, called photosynthetically active radiation (PAR). PPFD (photosynthetic photon flux density) is the amount of PAR reaching a surface each second, in µmol/m²/s. It is measured directly with a quantum (PAR) meter.

DLI is the daily total

DLI (daily light integral) is the total PAR reaching a surface over 24 hours, in mol/m²/day. Growth depends on this daily total, not on instantaneous brightness. ISU gives the formula:

DLI = PPFD × hours of light × 0.0036

The factor 0.0036 comes from multiplying by the 3,600 seconds in an hour and dividing by 1,000,000 to convert µmol to mol. If you use the 24-hour average PPFD, the factor becomes 0.0864 (86,400 seconds divided by 1,000,000), as Purdue HO-238-W explains.

In short: lux is brightness for people, PPFD is the instantaneous intensity of usable light, and DLI is its daily total. The table in the previous section uses DLI because daily totals are the right way to compare what plants need.

Estimating DLI from window lux, and where that breaks down

Even with only a light meter or a smartphone, you can roughly convert lux to PPFD. The conversion comes with conditions, though.

The conversion factor depends on the light source

Sensor maker Apogee Instruments shows in its conversion table that the lux to PPFD factor depends on the spectrum. Multiply lux by 0.0185 for sunlight, 0.0135 for cool white fluorescent and 0.0141 for metal halide. High pressure sodium varies by lamp type: 0.0122 for a mogul base lamp and 0.0130 for the double-ended product listed. For example, 108,000 lux of direct sunlight is 108,000 × 0.0185, or about 2,000 µmol/m²/s.

Use the sunlight factor of 0.0185 only when measuring natural light at a window. It does not apply where room lighting or a grow light mixes in. White LEDs are not in the Apogee table, so this article does not cover them.

A worked example under stated assumptions

The following numbers are hypothetical, meant only to show the orders of magnitude. They are not measurements.

Comparing steps 3 and 4, the lux you would need to hold for ten hours differs by roughly four times between the floor for foliage plants and the floor for leafy greens. A momentary 20,000 lux at the window does not mean it lasts ten hours.

One reading does not tell you the daily total

This calculation has a large limitation. Purdue HO-238-W states that a single measurement at one point in time does not accurately represent the light a plant received that day. Apogee also notes that when light changes over time, you cannot get DLI by multiplying by a constant factor. Window light swings throughout the day with sun angle, clouds and the shadows of nearby buildings.

For reference, Purdue cites outdoor DLI in the United States ranging from 5 to 60 mol/m²/day as monthly averages depending on latitude, season, day length and cloud cover, with northern latitudes at 5 to 30 from December through March. Those are outdoor US values, not indoor values behind glass. We could not confirm primary data on winter DLI in Japan, so we do not quote any.

So you cannot claim to have “calculated” DLI from one windowsill reading. What several readings can tell you is whether a spot is closer to the low-light group or might reach the high-light group. Pass that coarseness on to the AI as well.

If the numbers show your window will not reach the high-light group, add a grow light. A screw-in E26 bulb like this one fits an existing clamp lamp and suits one to a few pots; check the published PPFD at your hanging distance before you buy.

SANSI 32W Full Spectrum LED Grow Light Bulb (E26)

USD 26.36 on Amazon.com (as of 2026/09/29)

Amazon

＼楽天ポイント4倍セール！／

楽天市場

ポチップ

Measuring window light with a smartphone

Before buying a light meter, you can get a feel for your windowsill with a phone. How phones measure illuminance differs by platform, though.

Android and iPhone measure differently

On Android phones that have an ambient light sensor, apps read it through the Android sensor API as TYPE_LIGHT, in lux. The sensor is only available if the manufacturer built it in, and most manufacturers use it to adjust screen brightness. A research team at the University of Maribor that compared smartphone lux apps with a lux meter (DAAAM International Scientific Book 2022) reports that iOS apps measure illuminance through the camera instead.

The same paper offers a practical guideline. Among the ten phones tested (eight Android, two iPhone), Android readings were identical across apps because every app reads the same sensor, while iOS readings varied by app. That is a tendency, not a result proven for every phone and app. The authors also conclude that without proper calibration phones are not reliable for precise illuminance measurement, but they can provide meaningful data as a guide for home, photography and plant growing, and they cite earlier work showing larger errors in dim conditions.

In other words, treat phone readings as relative values for comparing places and times with the same phone and the same app, not as absolute numbers. On an iPhone, sticking with one app is especially important.

Measurement procedure

Keep these conditions the same to reduce scatter.

• Use the same phone and the same app. Switching either makes values impossible to compare.
• Measure at the same position and orientation. Place the phone at leaf height where the pot will sit, sensor facing up, and repeat the same way every time.
• Measure several times a day: at least morning, midday and late afternoon at the same spot. One reading is only a snapshot.
• Measure on both sunny and cloudy days, and record the cloudy values too.
• Vary the distance from the window: right behind the glass, 30 cm back, deeper in the room, one spot per candidate location.
• Note whether a sheer curtain was drawn, since that changes the reading.

The result is a table of lux by place and time. That table becomes the core of what you give the AI later.

General tendencies by window direction

Nebraska Extension notes that the brightest spot in a house is usually a south window and the darkest is along a north wall. East and west windows receive about the same amount of light, but west windows get considerably warmer. This applies to the Northern Hemisphere; in the Southern Hemisphere north windows are brightest. Writing down which way each window faces makes the readings easier to interpret.

Once your phone has shown the pattern, a dedicated lux meter makes repeated readings more consistent. It still measures lux, so converting to PPFD is valid only for natural window light, just as with a phone.

BTMETER BT-881D Digital Lux Meter (0-200,000 lux)

USD 39.99 on Amazon.com (as of 2026/09/29)

Amazon

＼楽天ポイント4倍セール！／

楽天市場

ポチップ

Winter temperature and humidity: the next filter after light

After light, check temperature and humidity. Winter homes are warm by day with heating, cold near the glass at night and dry. All three affect what to grow and where.

Temperature guidelines

Nebraska Extension says most foliage houseplants tolerate daytime temperatures of 65–75°F (about 18–24°C), with nights 10–15°F (about 5.5–8.3°C) cooler. Colorado State University Extension (CSU) groups indoor plants as cool at 40–50°F (about 4–10°C), intermediate at 60–75°F (about 16–24°C) and warm above 75°F (about 24°C).

Leafy greens and herbs need checking one by one. According to University of Illinois Extension, basil is sensitive to cold: low soil or air temperatures slow its growth, and 50°F (10°C) can cause blackened leaves. It can still be grown on a sunny windowsill through fall, winter and early spring. Seeds germinate and grow best when the medium is at least 70°F (about 21°C).

For lettuce, the Cornell CEA handbook gives greenhouse set points of 24°C day and 19°C night with 50–70% relative humidity. Those are settings for commercial hydroponic greenhouses, not recommendations for a home windowsill, but they indicate what leafy greens prefer.

On winter nights the area near the glass gets colder than the middle of the room. For cold-sensitive plants like basil, one option is to move them a little away from the window at night. How cold it gets at the glass varies by house, so the reliable approach is to put a thermometer where the pot sits and record the nighttime minimum.

Humidity guidelines

Nebraska Extension says most houseplants other than succulents are injured below 20% humidity and prefer 40–60%, and that winter indoor humidity can drop below 20%. CSU also notes that heated rooms are far drier than the native habitats of tropical plants.

Place one thermo-hygrometer and record daytime and nighttime temperature and humidity. As with light, keep several days of minimums and maximums rather than a single reading; it makes the conditions you give the AI more accurate.

Give the AI your growing conditions as a table

You now have lux by windowsill, the minimum and maximum room temperature, and humidity. Put them in a table, give it to a chat AI and ask for candidates. A table leads to fewer misread conditions than prose like “I have a south window and it gets a bit cold in winter,” and when you check the answer you can trace which condition each suggestion was based on.

Example columns

Use columns like these. The numbers are made up for illustration, not measurements.

location,window_direction,distance_from_window_cm,time,weather,curtain,lux,room_min_C,room_max_C,humidity_pct
living_window_A,south,0,10:00,sunny,none,18000,12,22,35
living_window_A,south,0,13:00,cloudy,none,6000,12,22,35
living_window_A,south,30,13:00,cloudy,none,3500,13,22,35
bedroom_window_B,east,0,09:00,sunny,sheer,4000,10,18,40
kitchen_counter_C,north,150,12:00,sunny,none,800,14,23,30

Add what you want to grow next to the measurements: “I want to eat leafy greens,” “I want to pick basil all winter,” “I want one foliage plant.” With wishes and conditions in one table, you are more likely to get answers of the form “this wish is hard to meet in this spot.”

Upload a file or paste into the prompt

There are two ways to hand over the table. One is to upload it as a CSV or similar file. The Claude Help Center says it handles CSV, TXT, JSON, XLSX and other formats (XLSX uploads require code execution and file creation to be enabled on the account). The OpenAI Help Center describes uploading spreadsheets such as CSV to have ChatGPT understand or visualize the data. The Gemini Apps Help says you can upload files including spreadsheets and, on the web, create charts from an uploaded table.

The other way is to paste a few rows like the example directly into the prompt. For a short table this is enough and works with any chat AI. Once you have measured for many days and the table grows, a file is easier to handle.

The table contains data about your home, such as window directions and room temperatures. How much of that to send to an outside service is your call. We discuss how to think about sending photos and household data to the cloud in our guide to working backward from local versus cloud AI, and our overview of the local AI stack is a starting point for what AI can run on your own GPU.

A prompt that asks for candidates and placement

Be explicit about what the AI should and should not produce. The goal is candidates with reasons, not a final decision. Here is an example.

The attached table lists lux, room temperature and humidity measured at my windows in winter.
The lux values are rough readings from a smartphone app, not a daily total.
For natural window light only, multiplying lux by 0.0185 gives an estimate of PPFD (µmol/m2/s).
List plants that could be grown indoors this winter under these conditions, for each location.
For each candidate, give:
1. The location (from the table) and distance from the window
2. The light the plant needs (DLI category) and whether the table conditions meet it
3. Whether the night minimum temperature and humidity are suitable
4. Possible fixes if conditions fall short (moving the pot, adding a grow light, etc.)
5. The type of source you based the judgment on
If a wish clearly cannot be met (for example leafy greens in the north spot), say so plainly.
Mark anything you filled in by guesswork as a guess.

The prompt has three deliberate features. First, it states up front that the lux values are momentary estimates; without that, the answer may treat a single reading as the daily light. Second, it limits the conversion factor to natural window light. Third, it asks the AI to say “not enough” when conditions fall short and to label guesses, which makes it clear what to verify.

Do not ask about diseases or pests at this stage. The purpose of planning is to narrow down plants and placements.

Check the AI answer against public horticulture references

Do not adopt the answer as is. Chat AIs can return wrong numbers or conditions in convincing prose. On a different topic, we recorded how the error rate of AI-generated exam questions changed sharply depending on whether we gave the AI source material (our test of AI-generated practice questions). Do not skip verification for growing advice either.

References to check against

Publications from US university Extension services (outreach programs run by state universities) are convenient because they present research-based numbers for a general audience. These are the ones used here and what each can verify.

How to check

Checking means comparing the answer with the references item by item.

1. Check plant groups. Find which ISU group each suggested plant belongs to. If the AI says lettuce or basil grows in low light, that contradicts ISU (high light).
2. Check the light estimate for each spot. Compare the work-back table above with your own lux table. If a leafy green is assigned to a spot that reads only a few thousand lux on cloudy days, doubt it.
3. Check the night minimum. If basil is assigned to a window that drops to around 10°C at night, revisit it in light of the Illinois cold damage note.
4. Check humidity. If your log shows days below 20%, foliage plants need extra humidity or a different spot.
5. Ask for sources of numbers. If the AI states a specific value such as “this plant needs a DLI of X,” ask where it comes from and see whether an Extension publication confirms it. Do not adopt numbers you cannot confirm.

When you find a mismatch, paste the relevant part of the reference into the chat and have the AI redo its candidates on that basis. With the reference in hand, it becomes easy to see where an answer drifts outside it.

Applying the references to basil, lettuce and foliage plants

Here is how the checking works for three plants that often come up. These apply reference values to hypothetical conditions; they are not growing results.

Basil

ISU places basil in high light (12–16). As the worked example showed, meeting that floor in ten hours takes about 18,000 lux of sunlight held for ten hours, which is hard on a cloudy winter day. Illinois Extension says windowsill growing in winter is possible but requires a sunny spot; basil also slows in the cold and can blacken at 10°C. A plan consistent with the references puts basil at the brightest point of a south window, moves it back from the glass on cold nights and adds supplemental light if needed.

Lettuce

The Cornell CEA handbook found DLI 17 best for the Boston bibb type it used, with 15 as the maximum for some cultivars before tipburn (dead leaf edges) appears. It also says growing at 17 requires strong downward airflow and that without it they could not exceed 12. So the ceiling depends on cultivar, spacing and airflow. These values come from commercial hydroponic greenhouses with supplemental lighting and fans, not from windowsills. Still, ISU also puts leafy greens in high or very high light, so growing lettuce on a winter windowsill alone is tough on light. If an AI says lettuce will grow in a north-facing kitchen, it conflicts with both references.

Foliage plants that tolerate low light

Nebraska Extension lists plants that tolerate low light, including Chinese evergreen (Aglaonema), cast iron plant, philodendron, pothos and sansevieria. ISU also places foliage plants in low (3–6) to medium light. Assigning these, rather than leafy greens, to north-facing spots and the back of the room is consistent with the references. We did not verify species-level DLI values, so judge by group and examples.

Put together, the references suggest this allocation within one home: the brightest point of a south window for leafy greens and herbs, east and west windows or spots farther from a south window for foliage plants, and the north side and back of the room for low-light tolerant foliage plants. If an AI answer departs sharply from this, ask it to explain how it read each condition, and judge against the references.

When you pick foliage plants for north-facing spots or the back of the room, a practical care guide organized by plant makes it easier to verify the candidates an AI suggests.

For the Love of Houseplants: Caring for and Keeping Plants with Confidence

USD 25.29 on Amazon.com (as of 2026/09/29)

Amazon

＼楽天ポイント4倍セール！／

楽天市場

ポチップ

Tools for measuring and adding light

Once a phone has shown you the pattern, here is how to think about tools for more accurate measurement or for adding light when it falls short. We cover types and use cases only, not prices.

Lux meters and quantum meters

A dedicated lux meter makes it easier to keep measuring conditions consistent than a phone app. It still measures lux, a different unit from usable light, just like a phone. For natural window light you can convert to a PPFD estimate with the sunlight factor, but you cannot make that conversion with confidence where artificial light mixes in.

To measure PPFD directly, use a quantum meter. The Apogee Instruments MQ-500 is one example. Quantum meters cost considerably more than lux meters, so for most readers they are the tool to consider if you want to measure PPFD seriously.

Grow lights

When window light is not enough, add a grow light. There are two broad forms.

• Screw-in spot lights with an E26 base: they fit existing clamp lamps and E26 sockets and suit supplementing one to a few pots on a windowsill.
• Panel lights (for example the Spider Farmer SF1000D or Mars Hydro TS 1000): meant to light a whole shelf or tent. They draw more power and run hotter, which is usually overkill for a few pots on a sill.

Under a grow light, the lux to PPFD factor depends on the spectrum, and the sunlight factor used in this article does not apply. Judge grow light output by the PPFD figures the manufacturer publishes. A screw-in bulb cannot plug into an outlet on its own; it goes into a fixture with a matching base. Some bulbs cannot be used in fixtures with a different base or in dimmable fixtures, so check that your clamp lamp qualifies before buying. Whatever the type, do not modify the wiring.

Thermo-hygrometers

Record temperature and humidity with a thermo-hygrometer placed where the pot sits. A model that logs minimum and maximum values makes nighttime cold at the window easy to see. Unlike light, temperature and humidity involve no unit conversion. Add a few days of records to the table and the conditions for the AI are complete.

A small thermo-hygrometer that keeps a history in a phone app lets you see several nights of window-side lows at once and copy them straight into the table you give the AI.

Govee H5075 Indoor Hygrometer Thermometer with Bluetooth App

USD 11.69 on Amazon.com (as of 2026/09/29)

Amazon

＼楽天ポイント4倍セール！／

楽天市場

ポチップ

Summary: measure, tabulate, ask for candidates, check against references

Planning winter indoor growing breaks down into four steps.

1. Measure: with a phone or light meter, record lux at each candidate spot in the morning, midday and afternoon, on sunny and cloudy days, under the same conditions. Log nighttime minimum temperature and humidity with a thermo-hygrometer.
2. Tabulate: one row per location, window direction, distance, time, weather, lux, temperature and humidity, plus what you want to grow.
3. Ask for candidates: give the table to a chat AI, state that lux readings are momentary estimates and when the conversion applies, and ask for plants per location with reasons.
4. Check against references: compare the answer item by item with the ISU DLI groups, Nebraska and CSU temperature and humidity guidance, Illinois on basil and Cornell on lettuce.

Two points matter most. First, lux is a unit for human eyes; what drives growth is DLI, the daily total of PPFD. You can estimate DLI from window lux, but you cannot calculate it from one reading. Second, when you use AI for growing advice, let the AI propose candidates and let public horticulture references decide.

Once your winter windowsill conditions are in a table, the table keeps working after planting. When leaf color or stem growth changes, comparing it with the light and temperature records for that spot helps separate low light from cold. Start by measuring your planned windowsill three times, morning, midday and afternoon.

Sources

• Iowa State University Extension: Important considerations for providing supplemental light to indoor plants
• Purdue Extension HO-238-W: Measuring Daily Light Integral in a Greenhouse
• Apogee Instruments: Conversion of PPFD to lux
• Apogee Instruments: Daily Light Integral
• Nebraska Extension NebGuide G2205: Guide to houseplants
• Colorado State University Extension PlantTalk 1317: Houseplants, temperature and humidity
• University of Illinois Extension: Basil
• Cornell CEA Hydroponic Lettuce Handbook
• Vujica Herzog et al.: Comparison between lux meter and smartphone apps (DAAAM 2022)
• Android Developers: Environment sensors
• Claude Help Center: Uploading files
• OpenAI Help Center: File uploads FAQ
• Gemini Apps Help: Upload files

ブラウザだけでできる本格的なAI画像生成【ConoHa AI Canvas】

swiftwand

X

AIを使って、毎日の生活をもっと快適にするアイデアや将来像を発信しています。 初心者にもわかりやすく、すぐに取り入れられる実践的な情報をお届けします。 Sharing ideas and visions for a better daily life with AI. Practical tips that anyone can start using right away.