Google Gemini

Google Gemini is no longer just a chatbot bolted onto Search — by late 2026 it is the consumer surface of Google's entire model stack, spanning text, images, video, music, code, and agentic tasks, with tiers from free to $199.99 a month. What interests me as a reviewer is the gap between the breadth on offer and the coherence of the experience. Few products in this category do more; few make it harder to know what you're allowed to do before you hit a limit. I have used every tier seriously over the past months, and my opinion of it depends almost entirely on which Google surface I'm standing on.

Where it sits

Gemini competes with ChatGPT and Claude for general assistant use, but it plays a home-turf game: personal context drawn from Gmail, Docs, Drive, and Photos; Gemini embedded in those same apps; Gemini Notebook (the research tool renamed from NotebookLM in July 2026) for source-grounded work; and AI Mode in Google Search for query-time reasoning. Google renamed its consumer tiers to Google AI Plus ($7.99/month), Google AI Pro ($19.99/month), and Google AI Ultra ($99.99/month for 5x Pro limits or $199.99/month for 20x), retiring the older Google One AI Premium packaging. The model lineup has churned fast: the app currently runs the Gemini 3.x family — Flash tiers for speed, 3.1 Pro for reasoning — with Deep Think reserved for Ultra and Gemini 3.5 Flash quietly powering the "Auto" selector.

For the Workspace-native user, the practical difference versus rivals is simple: the assistant shows up inside Gmail and Docs where the work already lives. For everyone else, the case rests on generation breadth and the free tier, which is unusually capable if you accept the quota.

What it actually does

The core app handles chat with long context (32,000 tokens free, 1 million tokens on Pro — Google frames that as roughly 50 pages versus 1,500), document and image understanding, Canvas for drafting, Gems for custom assistants, and Gemini Live for voice conversation. Media generation is unusually deep: Nano Banana 2 and Nano Banana Pro for images and editing, Veo 3.1 for video, Lyria for music tracks up to three minutes, and Flow credits for cinematic scene creation on higher tiers. Deep Research runs multi-minute investigations across "hundreds of searches" and returns cited reports; Deep Search does the same inside AI Mode. Screen automation (5 requests/day free, 20 on Pro, 120 on Ultra), Chrome auto-browse, scheduled actions, and the Gemini Agent on Ultra push the product toward task completion rather than pure conversation. Free-tier allowances are concrete if you dig: 20 images/day, 5 Deep Research reports/month, 10 music tracks/day, 20 slide presentations/day.

The subscription math has grown layered. Pro at $19.99 includes 4x free limits, 1M-token context, Veo 3.1 Fast video, Gemini in Gmail and Docs, 5 TB of storage, and expanded limits for the Jules coding agent and the Antigravity development platform. Ultra splits into two sizes now — $99.99 and $199.99 — adding Deep Think, Gemini Spark, Project Genie, 10,000 or 25,000 Flow credits, YouTube Premium, and Google Home Premium with camera-event AI descriptions. You can buy AI Credit top-ups on Pro ($25 for 2,500 credits) once you burn through the monthly allotment.

My testing notes

The most useful observation from my testing is how differently the same task behaves across tiers. On free, I ran a Deep Research request that returned a cited, multi-source report after a few minutes — genuinely comparable to paid rivals — but a follow-up conversation about a 60-page PDF ran into the 32k-token context wall almost immediately. On Pro, the 1M-token window swallowed a 200-page technical document in one pass and kept the thread coherent across a dozen follow-ups, and a spreadsheet plus three reference PDFs stayed in context simultaneously. That tier difference is where the subscription money actually goes: not a dramatically smarter model, but far fewer walls.

The second thing I watched was the metering. Google moved from daily prompt counts to compute-based limits that refresh every five hours against a weekly ceiling, where premium models and media generation "require more usage and may cause you to reach your limit faster." In practice this means a burst of Deep Research and video work drains the budget invisibly. I prefer knowing where I stand; Gemini currently makes that a guessing game, and the limits documentation spans several help pages that don't fully agree with each other. On the flip side, personal context is real and slightly uncanny: asking "what did the contractor say about the timeline" pulled the relevant thread from my connected mail without my naming the sender. It works, and it is exactly the feature that makes the privacy conversation non-abstract rather than theoretical.

What it does not do

Gemini's output quality is uneven across the model tiers — Flash answers are fast and sometimes shallow, Pro is strong but metered, and the Auto model switching hides which one actually answered you. It won't give you an API key inside the subscription; developer access bills separately through AI Studio or Vertex, so the consumer plan and the build plan are different budgets entirely. Some features are US-only or 18+ (parts of AI Mode, agentic features, Home Premium extras), and the fine print on what your conversations feed into Google's personalization data deserves a careful read before you point it at confidential material. Ultra's ceiling features — Deep Think, Project Genie, the highest agent limits — cost as much as a second broadband bill, and even Pro users report hitting weekly ceilings during heavy research weeks.

Where the ecosystem earns its keep

It is worth walking through what the surrounding products actually add, because the subscription is really a bundle. Gemini Notebook — still NotebookLM to most people until mid-2026 — is the piece I recommend most often to researchers: on Pro you get 500 notebooks with 300 sources each and 500 chat queries a day, and the audio overview feature turns a pile of PDFs into a listenable briefing in minutes. Jules, the asynchronous coding agent, and Antigravity, Google's agentic development platform, get expanded limits on Pro and the highest ceilings on Ultra; in my use Jules handles well-scoped refactors acceptably but is not replacing a hands-on developer for anything architectural. The bundling of YouTube Premium on Ultra and Google Home Premium with AI camera descriptions shows Google's instinct to fold every consumer service into the same funnel — useful if you already pay for those, noise if you don't.

A practical note for teams: the family sharing story is decent (up to six members on the consumer tiers), but the AI features carry age gates, so a household plan doesn't automatically mean every member gets the same assistant. And for buyers comparing against ChatGPT Plus or Claude Pro, the honest summary after my testing is this: Gemini wins on media generation and Workspace integration, trades blows on research via Deep Research, and loses on conversational consistency and metering transparency.

Verdict

Rated 4.3 on the ToolSpotter rubric: the functionality depth is category-leading and carries the 40% core-weighting — no competitor bundles text, image, video, music, research, and agentic tooling at this level — the free tier and $19.99 Pro plan are competitive on the 20% value slice, usability is the mixed part of the report card because of the opaque metering and model switching, the ecosystem across Workspace, Search, Android, and Chrome is unmatched on the 10% ecosystem slice, and reliability backed by Google's infrastructure covers the final 10%. For anyone already living in Gmail and Docs, this is the default assistant I recommend checking first. Just go in knowing that "free" means a 50-page context window, a per-day quota of the good stuff, and a set of usage rules that change more often than the models do.

Sources

• https://gemini.google/subscriptions
• https://9to5google.com/2026/05/25/google-ai-plus-pro-ultra-gemini-features
• https://9to5google.com/2026/04/11/google-ai-pro-ultra-features
• https://coursiv.io/blog/google-ai-pro-vs-ultra
• https://blogs.truescho.com/en/gemini-pro-vs-ultra
• https://support.google.com/gemini/answer/16275805
• https://blog.google/innovation-and-ai/products/gemini-notebook/notebooklm-gemini-notebook/
• https://one.google.com/ai
• https://gemini.google/
• https://play.google.com/store/apps/details?id=com.google.android.apps.bard