Image Alt Text: ChatGPT + Make

A small museum is digitizing its photo archive, thousands of images that screen-reader users cannot access because none have alt text. Writing a description for each one by hand is a job measured in months, so it never starts, and the archive stays closed to anyone who cannot see it.

This build writes the descriptions. Drop an image in a Drive folder, and ChatGPT Vision looks at it and writes concise, useful alt text plus a few tags, logged to a sheet. A backlog that was a wall becomes a queue that clears itself, one image at a time.

We wire it in Make, part of our AI image and vision set. Alt text is the corner where a vision model earns its keep: it reads what is in the picture, which is the part that takes a human real time, and does it at the scale a backlog needs.

What does this workflow actually do?

In one line: an image becomes its own alt text. Make watches a Drive folder. When you add an image, ChatGPT Vision describes it in a short alt-text sentence and a few tags, and Make logs both to a sheet. You copy the alt text into your site or catalog, accessible at last.

A few real cases, none of them the usual ones:

• The museum above, finally opening its archive to screen-reader users.
• An online store writing alt text for product photos, which helps both accessibility and search.
• A travel blogger captioning a years-deep backlog of trip photos.
• A nonprofit making its image library readable for everyone before a site relaunch.

Storing the images is done; they already exist. Looking at each one and describing it is the human time a backlog never gets. That looking is what a vision model does in a moment.

Why use Make if ChatGPT can describe an image itself?

Because a backlog is not one image, it is thousands. The ChatGPT app describes one picture you paste in, which is fine for a single image and hopeless for an archive.

Make turns it into a pipeline. It watches the folder, sends every image to the vision model, and logs each description without a click. The model does the seeing; Make does the doing-it-for-all-of-them and the filing. That division is what makes accessibility at scale actually happen instead of staying on the someday list.

What do you need before you start?

• A Make account. The free 1,000 operations a month covers a steady stream of images.
• An OpenAI API key from the OpenAI platform (the developer key, not the chat app). New to API keys? See how to use Claude for the idea.
• A Google account with a Drive folder of images and a Sheet to log into.
• Nothing else. The vision model reads the image straight from Drive.

Two Make words. A scenario is the whole folder-to-sheet automation. A module is one box. Three boxes, one scenario.

How does the workflow work, step by step?

Three modules, left to right:

Module	App	What it does
1. Trigger	Google Drive	Fires when you add an image to the folder
2. Brain	OpenAI Vision	Looks at the image and writes alt text and tags
3. Output	Google Sheets	Logs the alt text and tags as a row

Step 1: Watch an images folder

Create a scenario and add Google Drive, Watch Files in a Folder. Connect your account and choose the folder of images. This trigger fires on each new image and passes it to the vision step.

Step 2: Let OpenAI Vision describe it

Add OpenAI, Analyze images (Vision), paste your key, and map the image from Step 1. Tell it exactly what you want:

Write alt text for this image in one clear sentence, under 125
characters, describing only what is visible. Then list 5 short tags.
Do not guess brand names or text you cannot clearly read.

The “only what is visible” rule keeps the description accurate, and the length cap keeps the alt text usable on a real website.

Step 3: Log the alt text

Add Google Sheets, Add a Row, and map the image name, the alt text, and the tags into columns. Now you have a ready-to-paste accessibility sheet for your whole library.

How do you run it and check the result?

Click Run once on a test image. When I tested this, the alt text was usable right away, and the tweak that mattered was telling the model to describe only what is visible, so it stopped guessing at brand names and unreadable text. Tune the prompt, then turn the scenario on so the backlog clears itself.

After that, accessibility stops being a someday project. Every image that lands gets a clear description, and a library that was closed to screen-reader users opens up.

What does this cost to run?

Piece	Free tier	If you outgrow it
Make	1,000 operations/month free	Core plan from about $9/month
OpenAI Vision	Pay per image	Reading one image runs a fraction of a cent to a couple of cents
Google Drive	Free	Free
Google Sheets	Free	Free

Vision reads are cheap per image, though a huge archive adds up, so run it on the folders that matter first. More on tiers in our Make guide.

What can go wrong, and how do you avoid it?

• The alt text is too long. Cap the length in the prompt; good alt text is short.
• It guesses brand names or text. Tell it to describe only what is clearly visible, the fix that mattered most.
• Private images. Mind which folder you point it at; vision reads whatever is there.
• A giant archive runs up cost. Start with the images that are actually published, not everything at once.

The same see-then-log pattern runs many jobs. See Make AI scenarios.

How do you build this in Zapier or n8n instead?

Same three jobs, different names.

Job	Make	Zapier	n8n
Catch an image	Watch Files in a Folder	New File in Folder trigger	Google Drive Trigger node
Describe it	OpenAI Analyze images	OpenAI Vision action	OpenAI node
Log the text	Add a Row	Create Spreadsheet Row	Google Sheets node

Make and Zapier are easiest to start. Zapier vs Make vs n8n compares all three.

Common questions about this build

Is this good enough for real accessibility?

It is a strong first pass that gets a backlog described. Skim and fix the few that matter; that is far better than thousands with no alt text at all.

Does it help SEO too?

Yes. Descriptive alt text and tags help search engines understand images, so the same build serves accessibility and search.

Can it write captions instead of alt text?

Yes, just change the prompt. Alt text is short and literal; a caption can be longer and more stylistic.

Which model reads the image?

OpenAI’s vision model here. Gemini and others also do vision; swap the module if you prefer one.

Does Claude do vision on Make?

Not through a Make module today, which is why this build uses OpenAI Vision. Claude remains the pick for the writing builds.

Sources and official docs

• Creating a scenario (Make Help)
• OpenAI Vision (docs)
• Image alt text guidance (W3C WAI)
• OpenAI on Make (integrations)
• Alt text — Grokipedia

Last reviewed: May 2026. Make, OpenAI, and Google update their interfaces; check the official pages for exact button names.

You may also like

• AI Automation for Beginners (the hub)
• Draft WordPress posts from a spreadsheet with Claude
• Summarize new documents with Claude + Make
• Make.com Complete Beginner’s Guide
• How to use Claude AI
• Make AI Scenarios: automations that save hours
• The AI Tools Directory