Migrate from gemini to claude.
14 hands-on-verified translation patterns — what carries over and what to watch for. Cited to the Feature Parity Map; the audit tells you whether the move is worth it.
Claude Artifacts and Gemini Canvas are direct counterparts: a side panel that renders generated content (documents, code, React components, HTML, charts) iteratively as you prompt. Both products let you export to office formats — Claude downloads .docx/.pptx/.xlsx/.pdf inline, Gemini exports Canvas documents to Google Docs and slides to Google Slides. Claude added Live Artifacts (refreshable dashboards with MCP data sources like Notion, Slack, Sheets, Gmail) in April–May 2026; Gemini Canvas supports building runnable apps and visualizations and is unlimited slide-gen on AI Pro/Ultra (capped at 20/day on Free).
- Warning: Gemini Canvas exports natively to Google Docs/Slides (lives in Drive); Claude Artifacts download as files — there is no first-party 'send to Google Docs' button.
- Warning: Claude Artifacts have persistent storage up to 20MB per artifact and can be remixed by recipients with a Claude account; Gemini Canvas shares are snapshots without the same remix flow.
- Warning: Live Artifacts (Claude, 2026) pull real-time data via MCP connectors — there is no direct Gemini Canvas equivalent; Gemini relies on its own Connected Apps inside a fresh chat instead.
- Warning: Slide generation cap differs: Gemini Free caps at 20 presentations/day, Claude has no separate per-day cap for Artifact slides but is bound by overall message limits.
Both products run user/model-generated code in a sandbox to compute results, parse files, render charts, and iterate. Claude.ai's Code Execution runs Python and Bash in an isolated container (read/write only inside the working directory; network egress restricted to web search proxy), and the underlying tool (code_execution_20260120) supports REPL state persistence and programmatic tool calls on Opus 4.5+ and Sonnet 4.5+. Gemini executes code inside Canvas / Deep Think runs and surfaces results inline (charts, tables, generated files). Migration is essentially direct: data-analysis, chart-from-CSV, and one-shot script tasks port cleanly in either direction.
- Warning: Claude code execution requires the feature to be enabled on the account (also a prerequisite for Skills); some Team/Enterprise admins gate it off.
- Warning: Sandbox network rules differ — Claude's sandbox is locked down to a proxy + web search; Gemini's execution environment is similarly sandboxed but the exact policy is less publicly documented.
- Warning: State persistence: Claude code execution on Opus 4.5+/Sonnet 4.5+ keeps REPL state across tool calls in one conversation; replicating multi-step state-dependent work in Gemini may require keeping all steps in the same Canvas session.
- Warning: Neither sandbox is a substitute for Claude Code or a local IDE — long-running, multi-file, real-repo work should live in Claude Code, not in the chat sandbox of either product.
Claude.ai connectors are built on MCP — Anthropic ships first-party connectors for Google Drive, Gmail, Google Calendar, GitHub, Slack, and Microsoft 365 with login-only setup, plus a 375+ directory of third-party MCP servers (Linear, Notion, Stripe, Jira, Salesforce, Figma, etc.). Gemini's 'Connected Apps' is narrower and Google-owned: Workspace apps (Gmail, Drive, Docs, Calendar, Keep, Tasks) reached GA in May 2025, plus Google Maps, YouTube, Flights, and Hotels — no third-party MCP marketplace at the consumer-app level. Migrating from Claude to Gemini means you keep Google Workspace + a few Google verticals, but lose the long tail of SaaS connectors; replace them with Workspace exports or with Claude on the side.
- Warning: Gemini's Workspace connection is admin-gated on work/school accounts; same is true for Claude connectors on Team/Enterprise — verify admin policy before promising parity.
- Warning: Gemini removed '@Google Maps' and '@YouTube' as explicit @-mentions in October 2025 in favor of auto-routing; Claude connectors are still explicitly invoked through the connectors menu.
- Warning: Claude.ai's 375+ MCP connector directory has no Gemini analog — Salesforce, Linear, Stripe, Notion, Jira and similar SaaS tools will not be reachable from gemini.google.com (you'd need Gemini in Workspace add-ons or third-party glue).
- Warning: Both products have a Gmail/Drive/Calendar trio, but Gemini's are tighter (deeper Workspace integration) and Claude's are MCP-shaped (broader but more uniform).
Both products generate read-only public URLs for individual chats. Claude.ai: Share button on a chat creates a link anyone can open without a Claude account; account holders can fork/continue the conversation. Gemini: Share > Create public link generates a g.co/gemini/share/... URL; anyone can view, and signed-in viewers (excluding under-18s and Gem-based chats) can continue in their own Gemini Apps. On both, shared links are snapshots at link-creation time — later edits and new turns don't propagate.
- Warning: Gemini Gem-based chats and under-18 viewers cannot continue a shared chat; Claude allows any logged-in user to fork from a shared link.
- Warning: Work/school admin gating: Gemini Workspace admins can disable conversation sharing organization-wide (March 2026 announcement); Claude Team/Enterprise admins likewise control sharing — verify admin policy before relying on links.
- Warning: If Canvas content needs to be shared but public links are disabled by admin, export to Google Docs/Slides; Claude's equivalent is to download the Artifact as .docx/.pdf and share that.
- Warning: Neither product produces a 'team workspace' share — both are one-off public links, not collaborative real-time editing.
Anthropic turned on Memory for all Claude.ai users (Free + paid) in March 2026: Claude auto-summarizes conversations every ~24 hours, maintains a global memory plus a separate memory space per Project, and acknowledges in-chat when it recalls something. Gemini's equivalent is Apps Activity (chronological chat history at myactivity.google.com/product/gemini) plus a separate opt-in 'personalization from past chats' that lets a new chat reference prior conversations by topic or timeframe. When migrating, expect symmetry of capability but different controls: Claude memory is on by default with per-Project scoping; Gemini's cross-chat personalization is a separate opt-in toggle and is not scoped per-Gem.
- Warning: Gemini Apps Activity has explicit auto-delete windows (3/18/36 months, or never); Claude.ai's memory retention is governed by account settings rather than fixed timer buckets — review settings on both sides.
- Warning: Turning Apps Activity OFF on Gemini still keeps chats for 72 hours for abuse review; the equivalent retention floor on Claude depends on plan and may differ for Team/Enterprise.
- Warning: Per-Project memory on Claude is automatically isolated; on Gemini you must use separate Gems or Temporary Chats to get similar isolation.
- Warning: Both products have a 'temporary / incognito chat' mode that bypasses memory — Claude has incognito-style chats and Gemini has Temporary Chats; sensitive prompts should use these.
Gemini Gems are persistent, shareable custom assistants with name, instructions, and Knowledge files — directly analogous to OpenAI's GPTs. Claude.ai has no GPT/Gem-builder of its own: the closest combination is a Claude Project (for instructions + knowledge) plus an installed Skill (for reusable procedural workflow under Settings > Capabilities > Skills). For users coming FROM Gemini Gems TO Claude.ai, re-create each Gem as a Claude Project (paste Gem Instructions, re-upload Knowledge files) and, if the Gem encoded a structured workflow, package that as a Skill (SKILL.md + scripts) installed at the account level.
- Warning: There is no public Claude.ai gallery/marketplace of shareable assistants comparable to the Gemini Gem gallery; Skills are distributed as ZIPs or via Team/Enterprise org directories.
- Warning: A Gem is a single object users can switch to from the Gem picker; a Claude Project is a workspace you open into — the UX of 'pick an assistant per turn' does not exist in Claude.ai outside Skills auto-triggering.
- Warning: Premade Gems (Brainstormer, Career guide, Coding partner, Learning coach, Writing editor) have no 1:1 first-party Claude equivalents — re-create or replace with Skills.
- Warning: Skills require code execution to be enabled on the Claude account, which can be turned off by Team/Enterprise admins; Gems do not require any such toggle.
Both products have a multi-step research agent: Claude.ai's Research and Gemini's Deep Research. Each plans, browses many sources, iterates, and produces a long-form report with citations. On Claude, Research is paid-only (Pro, Max, Team, Enterprise) and requires Web Search on; on Gemini, Deep Research is available on all tiers with steep quotas (Free: 5/month, AI Plus: 12/day, AI Pro: 20/day, AI Ultra: 120/day) and Deep Research Max (Gemini 3.1 Pro, MCP support, richer visualizations) launched April 2026. Both can extend into your connected data — Claude searches Workspace (Gmail, Calendar, Docs) when those connectors are enabled; Gemini's Deep Research can optionally pull from Gmail, Drive, and Chat.
- Warning: Gemini Deep Research outputs include charts/infographics in 2026; Claude Research reports are primarily prose with citations (use Artifacts separately for charts).
- Warning: Quota model differs: Claude Research counts against your overall Pro/Max usage cap; Gemini Deep Research has explicit per-day caps per tier.
- Warning: Claude Research is unavailable on the Free plan; if a customer is on Claude Free and migrating to Gemini Free, they gain (limited) Deep Research access.
- Warning: Both bind into Workspace, but only Gemini Deep Research natively connects to Google Chat; for Slack-style team chat research, use Claude's Slack connector.
Both products accept rich file inputs in chat. Claude.ai: up to 20 files per conversation, 30 MB per file, PDFs up to ~32 MB or 100 pages for full visual+text analysis (larger files via the Files API up to 500 MB). Gemini: up to 10 files per prompt, 100 MB per non-video file, videos up to 2 GB; can also ingest one code folder or one GitHub repo per chat (up to 5,000 files / 100 MB). Migration is mostly direct — pick the product whose limits fit the workload: Gemini wins for video/audio and bulk code-folder ingestion; Claude wins for many-file conversations and per-page visual PDF analysis.
- Warning: Gemini natively ingests audio (up to 3 hours on Pro/Ultra) and video (up to 1 hour on Pro/Ultra) for transcription/QA; Claude.ai has no first-party audio or video understanding — fall back to transcribing first or use a connector.
- Warning: Claude can take 20 files per conversation; Gemini caps at 10 files per prompt — for many-attachment workflows, send multiple turns in Gemini.
- Warning: Both products' real limit is context window: Claude Sonnet 4.6 / Opus 4.6 have 1M token context, Gemini AI Pro/Ultra also 1M, Gemini Free is only 32K — heavy migrations from Claude Pro to Gemini Free will silently truncate.
- Warning: GitHub-repo and code-folder ingestion is a Gemini-specific affordance; the Claude equivalent is Claude Code on the local machine, not the Claude.ai chat UI.
On Claude.ai, account-wide standing instructions live at Settings > Profile (Preferences) and load into every new conversation. The closest Gemini analog has two separate fields: Settings > Personalization > Custom Instructions (default response style/format) plus Saved Info (long-term factual memory like 'I'm a vegetarian'). When migrating, split a Claude Preferences block into a 'how to respond' piece (paste into Gemini Custom Instructions) and a 'who I am / what I'm working on' piece (add as discrete Saved Info entries, or just tell Gemini 'remember that...').
- Warning: Gemini lets you add memories conversationally via 'remember X'; Claude.ai Preferences is a single Settings text box — there is no in-chat 'remember this' command (the closest is Claude's separate Memory feature).
- Warning: Claude.ai Preferences is one global field that applies to every chat outside Projects; Gemini Saved Info entries are individually toggleable list items you can edit/delete one by one.
- Warning: Project-specific instructions exist in both products (Claude Project instructions vs. per-Gem instructions), and they override or stack with the global ones differently — don't dump everything into the global box.
- Warning: Google warns against putting passwords, full addresses, or financial data into Saved Info; the same caution applies to Claude Preferences.
As of May 2026, Claude.ai consumer tiers are Free, Pro ($20/mo or $17/mo annual), Max 5x ($100/mo), and Max 20x ($200/mo); business tiers are Team (Standard $25/seat or Premium $125/seat, 5-seat minimum) and Enterprise (custom). Gemini's consumer tiers are Free, Google AI Plus ($9.99/mo), Google AI Pro ($19.99/mo with 2 TB Drive + family sharing), and Google AI Ultra ($249.99/mo, with reports of reduction toward $200 post-I/O 2026, 30 TB storage + Deep Think + Gemini Agent). Rough budget map: Claude Pro ~ Gemini AI Pro at $20; Claude Max 5x ($100) sits between AI Pro and AI Ultra; Claude Max 20x ($200) ~ rumored reduced AI Ultra price point.
- Warning: Gemini AI Pro bundles 2 TB Google One storage and family sharing for 5 — Claude Pro is just chat capability with no cloud storage included.
- Warning: Gemini AI Ultra includes Deep Think (192K reasoning) and Gemini Agent (up to 200 requests/day) — Claude Max has no agentic-runtime equivalent (Claude Managed Agents launched April 2026 but is API-side).
- Warning: Team pricing comparison: Claude Team Standard $25/seat vs. no exact Gemini consumer Team equivalent (Google's team product is Gemini for Workspace Business/Enterprise, sold via Workspace, not via gemini.google/subscriptions).
- Warning: Pricing is in flux in 2026 — verify the current Ultra price (reports of $200 vs. listed $249.99) and any new developer-focused Ultra tier before quoting; do not assume parity at any number.
A Claude.ai Project bundles per-project instructions, a knowledge base of uploaded files (PDFs, docs, text — RAG-expanded on paid plans), and the chats that share that context. Gemini has no fully-GA equivalent: a folder-style 'Projects' feature is in limited rollout as of May 2026 and not yet documented at the help-center level, so most users approximate with a Gem (instructions + Knowledge files attached to that Gem) plus the Apps Activity history for cross-chat recall. Migrate by re-creating each Claude Project as a Gem: paste the project instructions into the Gem's Instructions, upload the same knowledge files into the Gem's Knowledge section, and start chats from that Gem.
- Warning: Gem Knowledge files behave like attachments at chat time, not a persistent RAG index sized 'up to 10x context' the way Claude.ai's paid Project knowledge does; large knowledge bases may not all fit.
- Warning: Gemini's Projects rollout (folder-style chat organization) is reported but not yet on official Google help pages — do not promise customers feature parity with Claude Projects.
- Warning: Chats started from a Gem cannot be 'continued' by a recipient of a shared link, which is a stronger limitation than Claude.ai's shared Project chats.
- Warning: Cross-chat memory inside a Claude Project is automatic per-Project; on Gemini, cross-chat recall comes from Apps Activity + opt-in personalization-from-past-chats and is not scoped per-Gem.
This is the biggest asymmetry in the pair. Gemini natively generates images (Imagen) and videos (Veo 3.1 Fast / Veo 3.1, depending on tier) — Free gets 20 images/day, AI Pro gets 100 images/day + 3 Veo 3.1 Fast videos/day, AI Ultra gets 1000 images/day + 5 Veo 3.1 videos/day. Claude.ai does NOT generate raster images or video natively: it can take images as input (Vision), and via Artifacts it can produce SVGs, Mermaid diagrams, React charts, and HTML visualizations. For users migrating FROM Gemini TO Claude, raster image/video generation has no first-party replacement — either keep Gemini for those tasks, or wire a Stable Diffusion/FLUX MCP connector into Claude.
- Warning: Do not promise Claude image-gen parity with Imagen — Claude can only describe, analyze, or vectorize via SVG, not produce photorealistic raster images.
- Warning: Video understanding asymmetry too: Gemini ingests video natively, Claude does not (this is captured separately under file uploads).
- Warning: Gemini's image-gen quotas reset daily per tier; Claude has no equivalent quota because the capability doesn't exist.
- Warning: Music generation (Gemini: 10–20 tracks/day depending on tier) likewise has no Claude.ai analog.
Both products offer full spoken conversations. Claude.ai Voice Mode is available on every tier (Free included, with daily caps ~20–30 interactions); five voice options (Buttery, Airy, Mellow, Glassy, Rounded), 18 languages as of 2026, push-to-talk via space bar in Claude Code, and seamless switching between text and voice in the same conversation. Gemini Live is the equivalent on Google's side — real-time spoken conversation with similar multilingual support and built-in web search during voice turns. Migration is essentially feature-for-feature; on Gemini, Live additionally has camera/screen-share affordances in the mobile app that Claude Voice Mode does not match.
- Warning: Gemini Live has live camera and screen-share input on mobile (you can show it something while talking); Claude Voice Mode is audio-only.
- Warning: Free-tier voice caps exist on both; users running long voice sessions should be on paid tiers (Claude Pro/Max or Gemini AI Pro/Ultra).
- Warning: Voice + connectors: both can call web search mid-conversation, but neither has full parity with the typing UI for every connector — test multi-tool workflows.
- Warning: Claude voice in Claude Code uses push-to-talk (hold space); Claude.ai mobile voice uses a continuous-listening mode — UX differs by surface.
Both products ground answers against live web search and inline-cite sources. Claude.ai web search is a toggleable tool that the model invokes when a question needs fresh info, returning citations you can hover/click. Gemini grounds against Google Search when appropriate (no explicit toggle in many surfaces) and likewise renders inline citations. Migrating is essentially trivial at the chat level — turn the equivalent on if it's off, and expect slightly different citation density (Gemini tends to cite more aggressively since search is the native grounding mode).
- Warning: Claude requires Web Search to be enabled for Research mode to function at all; on Gemini, Deep Research bundles its own browsing and does not require flipping a separate switch.
- Warning: Search index differences: Gemini uses Google Search (best-in-class freshness for Google-indexed content); Claude uses Anthropic's web-search infrastructure, which can lag for very-recent or low-pagerank content.
- Warning: Free-tier rate limits are quite different — Claude Free has web search; Gemini Free has search grounding but is the most restrictive tier on Deep Research (5/month).
- Warning: Both products may answer without searching when they decide the model already knows — explicitly say 'search the web for...' if you need freshness.