Updates

When an AI model slows down or stops responding, the team hands the work to another model on its own and the task keeps going. Steps that run at the same time are spread across different models, so tasks finish faster at busy times too.

If you stopped a task or it got stuck on a step, pick it up with "Continue where it stopped". Finished steps stay as they are and the interrupted step continues from where it left off. Continuing doesn't use another task from your allowance.

We now name the AI providers we work with and where your content may go. On paid plans, your content is processed only by providers that don't allow training models on it. On the free plan, some requests may be processed on providers' free tiers; users in the European Economic Area, the United Kingdom or Switzerland, and users who haven't selected their country, are not routed to these tiers.

Instead of stopping at a report, the team can put the work into your own space: a document or spreadsheet in your Google Drive, or an image inside the task. Nothing happens without your approval; you first see a card showing what will be done, you approve or decline, and you can undo it afterwards. For the same kind of work you can choose "don't ask again". Everything it does is listed in Action history in your account settings.

If you work with several clients or projects, you can link your chats and team tasks to them. Pick the client and project from the "+" menu in the message box; your choice shows above the box and you remove it with one tap. Manage your clients in account settings. If you don't use clients, nothing changes.

Some image requests could produce a picture unrelated to the description. Images are now generated from a description that keeps the main elements of your request.

On the task page, long titles were squeezed into a one-word column on phones; they now use the full width with the buttons below. The client picker above the chat used to cover the first message; it moved into the "+" menu.

If you connect from the European Economic Area, the United Kingdom or Switzerland and hadn't chosen your country, account settings said Google Search was on. It now shows the real status and tells you when consent is needed.

We added new examples to the Guides section. Each one has a copy-ready example, how to fill it in, and a ready prompt so ADA AI can do the same job with your details.

Invite a friend with the personal link on your account page. When they finish their first team task, you both get 2 extra team tasks.

We've written copy-ready examples for documents people often look for: lesson plans, cover letters, business plans, SWOT analyses, content calendars and more. Each guide includes the example itself, how to fill it in, and a ready prompt so ADA AI can do the same job with your details. We keep adding new guides.

You can add ADA AI to your phone's home screen and open it like an app. When you start a team task, we can send you a notification when the report is ready, so you don't need to keep the page open. The notification never includes the task's content, only that it's done. You can turn notifications off in your account settings.

For anyone who wants to try before signing up, we added six small tools: cover letter, lesson plan and survey question generators, plus petition, birthday message and slogan tools in Turkish. Fill in a few fields and get one result. We don't store what you type. If you need something bigger, hand it to the team with the same details.

You no longer have to decide which mode to use. Type what you need, talk, or add a photo; Auto mode answers quick questions right away and hands bigger jobs to the team. If you'd rather it went the other way, switch it with one tap.

When you see a listing, an email or a screenshot on your phone, send it straight to ADA AI from the Share menu. It lands in the input box; nothing reaches us until you hit send.

No need to write the same task again every week. Turn a task into a routine and the team starts it on its own, daily, weekly or monthly, at the time you pick. When the report arrives you see what changed since last time. Routines use your plan's task allowance; if it's used up, that run is skipped and we tell you why. Pause, edit or delete them from the Routines page in the sidebar.

On video work the team can now do three things itself. It reviews your opening line for what it promises in the first three seconds and whether it creates curiosity, then writes five alternatives; it doesn't make up scores. It pulls Shorts- and Reels-ready moments, with timestamps, from your subtitle or transcript file (SRT, VTT, TXT). And it reads the CSV you export from YouTube Studio and summarises views, retention and click-through rate; those numbers are calculated from your file, not guessed by the model.

When you ask for a logo, the team draws three different concepts: each comes as an SVG file with an open-licence typeface suggestion, a colour palette and a small-size (favicon) version. These are starting concepts; they don't replace a trademark or similarity search.

If you use Claude, Cursor or another assistant that supports MCP, you can connect ADA AI to it. Create a personal access key in your account settings; your assistant can then hand tasks to the team, check a task's status, read the report or give work to a single specialist on your behalf. Tasks use your plan's allowance. You can delete the key at any time.

Answers in Fast mode now start almost immediately: simple messages go to a faster model, while anything that needs web search or long context goes to the stronger one. Team tasks that need only one specialist no longer get a separate report; you see the specialist's output directly, with the quality check warnings still in place. When Deep think is busy you no longer get an error; the answer comes from Fast mode and we say so under it.

Type "/" in the message box to open ready-made commands: /redteam for the strongest objections to an idea, /premortem to imagine why it failed six months from now, /swot, /hook10 for ten different opening lines, /trim to shorten text without changing its meaning, and /angles for five different takes on the same topic.

When you're not sure what to write, start from the templates page. There are ready-made tasks for every department: pick one, fill in a few fields and hand it to the team.

When you add a CSV or Excel file such as customer reviews or a request list to a chat, you can classify every row with the labels you choose. You download the labelled file and see the breakdown in a table; the numbers come from calculation, not the model. Before you start, we show how much of your allowance it will use.

Before you publish an idea, you can have it read through the eyes of a few different audiences: each gives its first reaction, its objection and what it finds convincing. These are estimates, not real research; the screen says so, and we never make up percentages or scores.

ADA AI now lives at getadaai.com. The old address, ai.adacreative.co, takes you to the same page on the new address automatically, so saved and shared links keep working. You may need to sign in once more after the move. Your account, chats, reports and brands are all still there.

Switch between Fast, Deep think and Delegate to team in one tap. The default is ADAM: it picks a suitable model for each message and moves to one that can read images when you attach one. You can also pick the model yourself; the ×1, ×2, ×3 next to each model shows how much of your daily allowance a message uses. Your brand profile is chosen from the same bar.

When a question needs current information, ADA AI searches the web and lists the sources it used under the answer. Which sources can be used depends on your country; you can set it in Account settings.

Ask for an image and ADA AI suggests an image description. Edit it as you like, press "Generate image" and the image is added to the chat, ready to download. The same button appears next to image descriptions in reports. A daily image allowance applies.

When you work on a long piece of text or code, edit it in the Canvas that opens next to the chat, compare changes and go back to an earlier version.

Attach a CSV or Excel file to a chat and ADA AI can analyse the data and give you tables and findings. The analysis runs in your browser; your data isn't processed on our servers. Files are kept for up to 30 days and you can delete them at any time.

In long chats ADA AI keeps a summary of the conversation so early details aren't lost. You can view and correct the summary from the chat screen.

An image attached to the first message of a new chat was sometimes not read; this is fixed. PDF attachments now work with any model you choose.

We now explain in detail how we protect your data (encryption, restricted access, deletion) and what is sent where when you generate images.

Paste a web page address into chat and ADA AI reads the page and bases its answer on it. Mention "SEO" and it checks the page title, meta description and heading structure. The source appears under the answer. A daily reading limit applies.

If you have connected your Google account under Account > Connections, chat and team tasks can work from your Search Console and Google Analytics data. Just mention "Search Console" or "Google Analytics" in your question. If you have a single site or property it is picked automatically; if you have several, name the domain in your question or set a default under Connections. Data is read only when you ask.

Answers now have copy, read aloud, good or bad response and retry buttons. Click the chat title to rename it, or delete it from the ⋯ menu. Read aloud uses your browser's voices.

Brands and the single-specialist link moved to the top. You can filter your history by All, Tasks and Chats. Long briefs show up with short titles in the list.

The permanent usage card in the sidebar is gone. When you reach your limit, a notice explains that you can wait for it to reset or upgrade your plan.

Report buttons are now grouped into editing, exporting and moving to chat. On phones, export options are gathered in a single menu.

The privacy policy now states that Search Console and Google Analytics data can also be used in chat and is sent to the AI model provider only within that request, to generate the answer.

Add a web page, an App Store app or a topic discussed on Hacker News to your brief and the team reads it and works it in. It can also check a page's SEO. The sources it read appear on the team board and in the report.

Edit the team's report right on the page. Every change is saved as a separate version, and you can go back to the original at any time. Numbers without a source stay flagged after you edit.

Download the report as a Word file and share it with your team. Anyone with the share link can download it too; that file leaves out the specialists' working notes.

When the team reads a page, it never follows instructions written in it; it only uses it as information. What it takes from a page is not treated as verified and is always written with its source.

On some new free accounts, chat, specialists and team tasks couldn't be used. This is fixed; the free plan is active on new accounts from the start.

The "Plan & Billing" link in notices now takes you straight to the plan tab.

The team has grown to 20 departments: software, legal, finance, games, security, design, health, education and more. Just write your brief; the planner picks the right specialists for the job. You can also work with a single specialist directly from the Specialists page.

See who works in each department on the new Departments page. Pick a sample brief and watch the team come together.

Every specialist now has their own pixel character, so you can see who's working on the team board at a glance. Each department has a lead.

Specialist names and descriptions are in Turkish and English. Switch the interface language and the specialist cards switch with it.

Use search in the sidebar or Ctrl+K to look through your chats, reports, brands and attachments. It finds things by topic, even when the words differ. Pick "Add this report as context" to bring an old report into a new chat or task; the reply shows which sources it used.

Add images, screenshots or PDFs to chats and team tasks, and the team reads them as part of the work. On the Brands page, "Fill from brand guide" pulls the details from a PDF into your brand; you check every field before saving. The file itself is deleted once it's been read.

Press the microphone in the chat box or the task form and talk; what you say is written into the box. You can edit it before sending. We don't keep the recording.

Messages and tasks now pass an automatic safety check before they reach a model. Illegal or harmful requests aren't sent and don't use up your quota. The text that's checked isn't stored.

If the service briefly stops responding, chat and team task screens tell you what happened and what to do. Messages that fail during that time don't count toward your chat quota.

Tick off action items as you finish them; your progress shows on every device. When a report has a calendar, download it as an .ics file, pick a start date, and add it to Google Calendar, Apple Calendar or Outlook.

The free plan includes 4 team tasks a week and the Team plan 60 a month. Chat now has its own daily quota, so a long team task doesn't use up your chat. You can see what's left in the sidebar.

A brief template you can fill in without signing in. Copy it, download it, or send it to the team in one click; after you sign up, your brief is waiting in the task form.

If you like, fill in the brief field by field: audience, budget, timing, deliverables, facts you know and things to avoid. There are also eight ready templates; pick one and replace the bracketed parts with your own details.

Describe your brand once: what you sell, who to, your tone, the phrases you never use and the facts you know. Pick the brand when you start a task or chat, and the team works with all of it from the first step. If a banned phrase shows up, the Editor sends the step back; facts you add count as sources and don't get flagged. 1 brand on the free plan, 10 on paid plans.

ADA AI now comes in three plans: Free (4 team tasks a week), Team (we run the models) and Bring Your Own Key (your provider, no task limit from us). Team is $12 a month and Bring Your Own Key $2 a month for your first 3 months. From October 23, using your own API key needs the Bring Your Own Key or Team plan. Cancel anytime from Account → Plan & Billing.

Registration is open. Create your account and put the team to work without waiting for an invite.

If your brief asks for something the team can't do itself, like launching an ad or sending an email, the report says so and explains how you can do it.

When you reach a daily task or usage limit, you see what happened and what you can do next.

We fixed a bug that could make letters appear in reverse order while you typed.

See how many team tasks you have left today and when they reset. No more token counts.

Account, team and specialist pages have English addresses. Your old links and bookmarks still work.

Rewritten for paid plans: payments, cancellation and where your data is processed.

Write your brief in Turkish, get a Turkish report; write it in English, get English. Your interface language no longer decides.

Ask for a content calendar, a slogan or a window display idea, and the planner gives each one to the specialist who does that job.

If a deliverable you asked for couldn't be produced, the summary tells you plainly.

If a plan goes over the budget in your brief, the editor flags it and the final report brings it back within budget.

If one specialist can't finish, the task doesn't stop. The report comes with the completed steps and says what's missing.

We fixed the timeouts that could leave long tasks hanging.

The team member who checks every step is now called the Editor. Same job: checks the work against your brief and sends it back if something's off.

Meeting lengths, calculated scores and splits that add up to 100% are no longer flagged as unsourced numbers.

Sending the same brief twice within a minute won't start a second task.

• Your new-chat draft is kept.
• Team tasks have a separate context field.
• The task view is wider with a single scrollbar.

Your account opens the moment you verify your email. No waiting for approval.

Share any report with a read-only link. It's hidden from search engines and you can turn it off any time.

Actions like delete now use the app's own confirmation and notification windows instead of browser pop-ups.

Sign-ups with disposable email addresses are blocked, and new accounts have extra limits in their first days to prevent abuse.

Write a brief: the planner splits the work, specialists do it, the editor checks every step, and you get one report.

Remove tasks you no longer need from the list or the task view.