Google's AI is no longer one product you open. It is twenty-two of them, spread across search, documents, images, video, code and music, and AI Central's 22 Best Google AI Tools names every one and what it is for. The practical read: Gemini is the front door, NotebookLM and Deep Research do the reading, a bank of image and video models does the making, and a newer tier of agents now writes and ships software. Pick three, not twenty-two.
Reviewed August 7, 2026.
Why this is a map, not a menu
AI Central's list is not a ranking. It is a survey of surface area, and the surface is the story. Gemini sits at the centre as the conversational assistant, but Canvas lives inside Gemini, Gemini in Workspace lives inside Gmail, Sheets, Slides, Photos and Maps, Imagen 4 is reached through Google AI Studio and the Gemini API, and Vids sits inside Workspace too. The rest are standalone destinations you visit on purpose.
That is why most people use two of these and miss twenty. The tools are not hidden, they are in five different places, and nothing in an ordinary working day points at any of them. AI Central frames the whole set around four kinds of work instead.
generate marketing visuals, automate research, code with AI assistance, or create professional videos from text
Those four verbs are a better organising principle than the numbering. Sorted by job rather than by the order they appear in, AI Central's twenty-two fall into five clusters. That grouping is a reading of the list, not its own structure, which runs as a straight numbered sequence.
- Ask and think. Gemini, AI Mode for search inside Google, Visual Layout for answers delivered in an immersive magazine style with photos and interactive modules, and Guided Learning, an AI tutor that paces complex topics to you.
- Read and research. NotebookLM for documents you supply, Deep Research for material it goes and finds.
- Make images, video and sound. Nano Banana Pro, Imagen 4, Veo 3.1, Flow, Vids, Whisk, MusicFX, and Canvas for decks, sites and documents built with Gemini in real time.
- Build software. Jules, Antigravity, Opal and Stitch, plus Gems and Google AI Studio for custom assistants.
- Work where you already are. Gemini in Workspace, drafting emails and designing slides inside the apps already open, and Pomelli, which reads your website to build a Business DNA profile and generates on brand campaign ideas from it.
The research pair is the part most people underuse
NotebookLM and Deep Research look like the same tool and are not. The difference is where the evidence comes from, and that single distinction decides which one you reach for.
NotebookLM works on a closed set. You upload the material, and per AI Central's description it returns mind maps, audio summaries, video overviews and study guides built from what you gave it. Nothing outside your files enters the answer. That is the point: a contract, a board pack, ten PDFs you have to actually know by Thursday.
Deep Research works on the open set, gathering, analysing and synthesising online information into polished, in depth documents. You are not interrogating a corpus, you are commissioning one.
So: if you already have the sources, NotebookLM. If finding the sources is the job, Deep Research. Using the wrong one is the most common way people conclude these tools do not work. AI Mode is the shallow end of the same behaviour, search inside Google returning summarised, source backed answers.
The creative stack is four decisions, not one
Eight of the twenty-two make something you look at or listen to, which is where AI Central's list is most crowded. For stills, Nano Banana Pro is the top tier model for viral ready marketing, ads, social and sales collateral, while Imagen 4 is the newer text to image model called out for photorealism and precise text rendering through AI Studio and the Gemini API. One is the creative workhorse, the other is what you address programmatically when the text inside the image has to be correct.
For moving image there are three ambitions. Veo 3.1 produces realistic video with synchronised sound. Flow turns text and media inputs into coherent cinematic scenes and narratives. Vids sits inside Workspace with templates, AI scripts and voiceovers, for the business video a colleague has to edit later.
Then the specialists. Whisk remixes images you upload as the prompt itself, using Gemini and Imagen 3. MusicFX composes original tracks from a description of genre, mood and instruments. Canvas is the one to try first, because it is already inside Gemini and turns a conversation into a presentation, a website, an infographic or a document without leaving the chat.
The agent tier is where the list gets ambitious
Six entries are not assistants, they are builders, and AI Central's framing for them is noticeably more aggressive than for the rest of the list. Jules connects to GitHub, clones your repository and works the codebase directly, handling bug fixes, tests, updates and pull requests. The important qualifier is that the output goes back for developer review. It proposes, a human merges. Antigravity removes that boundary, as an agent first development environment where the agents own the loop.
autonomous AI agents plan, write, test, debug, and verify software across the editor, terminal, and browser
Editor, terminal and browser is the entire surface a developer touches. That is the distance between an assistant that finishes your line and a system that runs the cycle.
Below that sit two no code paths. Opal builds AI powered mini applications from a plain language description and lets you edit and share them. Stitch converts a text or image prompt into complete web and mobile interfaces, with exportable front end code and a Figma integration.
One honest flag. AI Central describes Google AI Studio in exactly the same words it uses for Gems, custom assistants built without coding or chosen from a library of prebuilt niche assistants. On that wording the two are not distinguished, so treat them as one capability and test both on a single real task before standardising your team on either.
Five of the twenty-two are labelled experimental
Counting by the wording of the descriptions themselves, five carry the word experimental: Visual Layout, Pomelli, Stitch, Whisk and MusicFX. Roughly one in four. That count comes from reading each description and looking for the term, so it reflects how AI Central labelled them rather than any official product status.
Treat that as a planning instruction rather than a warning. Experimental is where the genuinely new capability shows up first, and Pomelli is the clearest case, since nothing else here reads your own website and turns it into a reusable brand profile. Use them, just not on the critical path of a client deliverable without a fallback.
What to actually do this week
Twenty-two is a browsing number. Three is the adoption number, and the right three depend on the work.
- Starting from nothing: Gemini for daily thinking, NotebookLM for anything you are obliged to read carefully, Gemini in Workspace so the help appears in Gmail and Sheets rather than in another tab.
- Marketing and content: Nano Banana Pro or Imagen 4 for stills, Veo 3.1 when the asset needs sound, Pomelli to get campaign ideas out of your own site instead of a blank prompt.
- Engineering: Jules for scoped repository work that ends in a reviewable pull request, Antigravity when you want the agent to run the whole loop, Stitch to skip the first draft of an interface.
- Teaching or learning: Guided Learning for pacing, NotebookLM to turn source material into study guides and audio you can listen to on the move.
AI Central closes with the only instruction that matters here.
You've got the tools Now, put them to work
The failure mode with a list this long is treating it as reading. Open three before the end of the week, give each a real task rather than a test question, and drop the two that do not survive contact with your actual work.
What is the difference between NotebookLM and Deep Research?
NotebookLM works only on material you upload, turning it into mind maps, audio summaries, video overviews and study guides. Deep Research goes out to online sources and synthesises them into an in depth report. If you have the documents already, NotebookLM. If finding them is the work, Deep Research.
Which Google tool should I use to make a video?
AI Central names three, for three different jobs. Veo 3.1 is the most advanced generator, producing realistic video with synchronised sound for campaigns, tutorials and promotional content. Flow builds coherent cinematic scenes and narratives from text and media inputs. Vids sits inside Google Workspace and makes collaborative business video from prompts, templates, AI scripts and voiceovers.
What is the difference between Gems and Google AI Studio?
AI Central gives both the same description, custom AI assistants built without coding, or a library of prebuilt assistants for specific tasks and industries. On that wording nothing separates them, so do not choose on paper. Build the same assistant in each, run it against one real task, and keep whichever your team opens again.