Continue Dropped 13.4 Points in September: What Developers Should Know

Continue is an open-source coding assistant that runs inside VS Code and JetBrains. Its share of AI recommendations in claw.mobile's index fell from 17.9% in August 2026 to 4.5% in September, a decline of 13.4 percentage points. This is the largest single-tool movement in the September snapshot and signals a meaningful shift in how general-purpose AIs answer tool-selection questions.

The practical signal is not that Continue became unusable. Continue remains actively maintained, technically capable, and highly configurable. The drop reflects how AIs weight visibility, workflow completeness, and answerability when choosing tools to recommend.

Continue's strength is control: developers can configure models, prompts, context windows, hosting, and repository workflows without accepting a vendor's defaults. That flexibility requires technical judgment, which makes Continue harder to recommend in broad "best for developers" queries.

The live scoreboard tracking all tool movements is available at claw.mobile/ai-recommendations, which compares recommendations from ChatGPT, Claude, Perplexity, and Grok across 40 real tool-choice questions.

Cursor, Windsurf, GitHub Copilot, Claude Code, Bolt, and v0 generally present a clearer first-run experience with integrated deployment, polished onboarding, and visible product marketing. Continue can still be the better technical choice, but it requires users to evaluate tradeoffs rather than following a recommended default path.

Partly. The recommendation drop is a useful warning about Continue's positioning, not proof that Continue is worse for developers. The AIs are more accurate for developers who prioritize speed and integration; they are less accurate for teams that value provider independence and deep configuration.

Continue remains compelling when the priority is flexibility. Use it when you need to run a preferred model or provider instead of accepting a vendor-controlled default, execute local or self-hosted models for privacy or cost control, customize rules and prompts for repository-specific workflows, keep the assistant integrated with an existing VS Code or JetBrains environment, or avoid moving projects into a separate hosted app-builder platform.

Cursor's integrated editor, agent, chat, and tab-completion workflow is easier to explain and easier to recommend broadly. Current pricing lists Free at $0 with limited AI usage, Plus at $8 per month, Premium at $30, Super at $100, and Ultra at $200, with usage pools that refresh weekly. Cursor's polish and clear feature tiers make it a safer general recommendation.

The AIs are less likely to be right for teams that value model governance, local inference, or deep customization. Continue's lower recommendation share may therefore reflect answerability and product visibility as much as product quality. The question "best for developers" rewards tools with clear deployment paths and minimal setup, which is a subset of what developers actually need.

Developers choosing a default tool should start with Cursor, Windsurf, GitHub Copilot, or Claude Code if the goal is to reduce setup friction and get agentic coding working quickly. Test Continue when model choice, local execution, or customization matters more than convenience.

Teams standardizing an AI workflow should not choose from the "best developer tool" label alone. Compare repository indexing depth, terminal access, code review integration, security controls, model governance policies, usage limits, billing transparency, and portability. Continue's configuration surface is a feature for teams that want control and a barrier for teams that want defaults.

Non-coders and app builders should not start with Continue. Glide, Adalo, Bolt, v0, and Replit align better with visual workflows, natural-language generation, templates, and deployment. Continue is designed for developers working inside code editors, not for users describing apps in natural language and receiving polished outputs.

Continue's maintainers and advocates face a positioning problem. "Open-source coding assistant" is technically accurate but less actionable than explicit use cases such as local-model development, provider control, privacy-sensitive repositories, and customizable IDE automation. The September result suggests that broad developer-tool questions favor integrated products over flexible assistants.

The AI recommendation market is rewarding complete workflows, not merely capable assistants. Tools that offer a clear path from prompt to deployed application receive stronger recommendations than tools that require configuration decisions, even when the configurable tool is technically superior for specific use cases.

This pattern appears across categories. Cursor's integrated environment beats Continue's flexibility in broad recommendations. Bolt and v0 win landing-page queries over lower-level frameworks. Glide's visual builder receives more recommendations than Firebase and Supabase for non-coders, even though Firebase offers more control.

Developers should treat Continue as a deliberate choice for control and extensibility, not the default answer to every coding-assistant question. The tool performs well when the user knows why they are choosing it: model governance, local execution, custom context rules, or avoiding vendor defaults.

The practical conclusion is that AI recommendations favor products with visible onboarding, integrated deployment, clear pricing, and minimal decision fatigue. Continue's recommendation share will likely remain low in broad queries and higher in specific searches for open-source, local-first, or vendor-independent coding assistants.

Continue's primary strength is provider independence. Developers can connect to OpenAI, Anthropic, Google, local models via Ollama, self-hosted endpoints, or Azure OpenAI without accepting a single vendor's model choices or billing structure. This matters for teams with model governance requirements, cost sensitivity, or privacy constraints.

The second strength is customization depth. Continue allows users to define custom prompts, adjust context windows, configure repository indexing rules, and integrate with existing CI/CD workflows. This level of control is useful for large codebases, domain-specific languages, or teams with established coding standards.

The primary weakness is setup friction. Continue requires users to choose models, configure API keys, and tune context settings before achieving the same out-of-box productivity that Cursor or GitHub Copilot provide immediately. This friction is acceptable for teams that value control; it is a barrier for developers who want to start coding within minutes.

The second weakness is visibility and product marketing. Continue is open-source and community-driven, which limits its presence in commercial tool comparisons, sponsored content, and vendor marketing. Lower visibility translates to fewer AI recommendations, even when the tool is technically competitive.

Choose Continue when model choice matters. If your team prefers Claude Opus for reasoning, GPT-4 for generation, and local models for cost control, Continue allows you to switch providers without changing tools. Cursor and GitHub Copilot lock you into vendor-controlled model defaults.

Choose Continue when local execution is required. Teams working on proprietary codebases, regulated industries, or air-gapped environments can run Continue with Ollama, llama.cpp, or self-hosted endpoints. Cloud-only tools cannot meet these requirements.

Choose Continue when you need deep customization. If your repository uses non-standard languages, domain-specific frameworks, or custom build systems, Continue's prompt and context configuration offers more control than integrated tools with fixed workflows.

Choose Cursor or Windsurf when speed and integration matter more than flexibility. If the goal is to start coding immediately, accept recommended defaults, and deploy quickly, Continue's configuration surface is overhead rather than value. For a detailed comparison, see Continue vs Cursor vs Windsurf.

The recommendation drop clarifies Continue's position: it is the right tool for teams that know why they need it, not the default recommendation for every developer. Evaluate based on your team's control requirements, not on aggregate recommendation share.

The September data confirms that AI recommendations reflect product visibility and workflow completeness as much as technical capability. Continue's 13.4-point drop does not mean it became worse; it means other tools present clearer answers to broad queries.

Developers should use AI recommendations as a starting point, not a final decision. Test the recommended tool, but also evaluate tools that offer different tradeoffs. Continue's low recommendation share does not disqualify it for teams that value model choice, local execution, or customization.

Track the live recommendation index at claw.mobile/ai-recommendations to see how tool recommendations shift over time. Month-over-month changes reveal positioning trends, not just product quality. Continue's drop, Cursor's decline, and the rise of integrated builders like Bolt and v0 all reflect how AIs weight different product attributes.

For related tool comparisons, see 9 Best Cursor Alternatives in 2026 and Best AI Coding Tool According to Reddit. For broader context on coding-assistant selection, see Cursor Dropped 11.6 Points in September.

The practical takeaway: treat AI recommendations as signals about positioning, answerability, and visibility. Evaluate tools based on your specific requirements, not on aggregate recommendation share.