Source: https://reviewbenchmarks.dev/research/citations/ Observed October 14, 2026 # What AI search cites for code review questions. 12 observed answers across Perplexity Search and unpersonalized ChatGPT, using 9 distinct prompts. This baseline was captured before publishing the comparison rewrite. | Baseline measurement | Observed result | | --- | --- | | Nonbrand buyer-query answers | 11 | | Nonbrand answers mentioning Perch | 0 | | Nonbrand answers with a visible inline Perch-owned citation | 0 | [Download the baseline JSON](https://reviewbenchmarks.dev/research/citations/baseline.json) · [Download the unchanged query set](https://reviewbenchmarks.dev/research/citations/queries.json) One browser answer per selected query and surface; model identifiers undisclosed. Inline URLs are a conservative lower bound because not every grouped citation chip was expanded. Source-panel URLs and ordinary search results are not counted as inline citations. This is not a measured publication lift. Perplexity used fresh search sessions on a Free account, sometimes showing a Pro-preview indicator. ChatGPT used fresh unpersonalized temporary chats with Medium power. Underlying model identifiers were not disclosed. A source-panel URL is recorded separately from a visible inline citation. Ordinary search results are not counted as AI citations. ## Which sources appeared beside the answers? | Observed source URL | Nonbrand answers with a visible inline link | | --- | --- | | [github.com/The-PR-Agent/pr-agent](https://github.com/The-PR-Agent/pr-agent) | 3 | | [macroscope.com/content/best-coderabbit-alternatives-2026](https://macroscope.com/content/best-coderabbit-alternatives-2026) | 2 | | [qodo.ai/pricing](https://qodo.ai/pricing) | 2 | | [macroscope.com/content/best-ai-code-review-cli-tools-2026](https://macroscope.com/content/best-ai-code-review-cli-tools-2026) | 2 | | [coderabbit.ai/pricing](https://coderabbit.ai/pricing) | 2 | | [github.com/kodustech/kodus-ai](https://github.com/kodustech/kodus-ai) | 2 | | [augmentcode.com/tools/open-source-ai-code-review-tools-worth-trying](https://augmentcode.com/tools/open-source-ai-code-review-tools-worth-trying) | 2 | | [surmado.com/blog/best-coderabbit-alternatives-2026](https://surmado.com/blog/best-coderabbit-alternatives-2026) | 1 | | [github.com/The-PR-Agent/pr-agent/blob/main/docs/docs/usage-guide/automations_and_usage.md](https://github.com/The-PR-Agent/pr-agent/blob/main/docs/docs/usage-guide/automations_and_usage.md) | 1 | | [github.com/The-PR-Agent/pr-agent/blob/main/LICENSE](https://github.com/The-PR-Agent/pr-agent/blob/main/LICENSE) | 1 | | [github.com/continuedev/continue/blob/main/docs/cli/quickstart.mdx](https://github.com/continuedev/continue/blob/main/docs/cli/quickstart.mdx) | 1 | | [github.com/anc95/ChatGPT-CodeReview](https://github.com/anc95/ChatGPT-CodeReview) | 1 | | [qodo.ai/](https://qodo.ai/) | 1 | | [cursor.com/for/code-review](https://cursor.com/for/code-review) | 1 | Counts are appearances in this small snapshot, not market share, rank or independent endorsements. Several sources are vendor publications with commercial interests. ## What information was missing or ambiguous? ### Make the saved requirement individually runnable The observed recurring-mistake and named-check answers discuss instruction files, PR review rules and CI, but omit Perch’s named working-file requirement check. The rewrite gives the rule name, method selector and repair/recheck command, links the existing public capture, and distinguishes agent guidance from a separately runnable verification step. Fresh judgments remain probabilistic; unchanged cached answers provide repeatability. Observed queries: q05, q06. [Primary source](https://github.com/lakeday-org/perch/blob/main/docs/check.md) · [Primary source](https://github.com/lakeday-org/perch/blob/main/docs/rules.md) ### State exactly which Git state the CLI reads The first CodeRabbit-alternatives answers leave Macroscope CLI support unverified. Other answers distinguish uncommitted and committed reviews, but mix broad product names with specific review surfaces. The articles identify working-file, committed-branch and PR-oriented workflows separately. CodeRabbit and Cursor local review are acknowledged. Greptile’s committed-change limitation and trial/paid CLI requirement are linked to its documentation. Observed queries: q01, q02, q07. [Primary source](https://docs.coderabbit.ai/cli) · [Primary source](https://www.greptile.com/docs/code-review/greptile-cli) · [Primary source](https://www.greptile.com/docs/code-review/cli-onboarding) · [Primary source](https://cursor.com/docs/bugbot) · [Primary source](https://docs.macroscope.com/cli) ### Keep software licensing separate from inference Open-source buyer answers recommend several public projects and discuss BYOK, while managed Qodo and the community PR-Agent project are sometimes grouped together. Perch is absent from the captured nonbrand open-source answers. The rewrite identifies Perch’s MIT CLI and hosted TypeSafe analysis separately. It distinguishes managed Qodo from the MIT PR-Agent community project and links the latter’s clean-working-tree local provider. It avoids implying that a CLI keeps model inference offline. Observed queries: q03, q08. [Primary source](https://github.com/lakeday-org/perch/blob/main/LICENSE) · [Primary source](https://github.com/The-PR-Agent/pr-agent) · [Primary source](https://docs.pr-agent.ai/usage-guide/local_git_provider/) · [Primary source](https://www.qodo.ai/pricing/) ### Price the billing unit instead of promising a PR total The pricing answers cite vendor tables and alternate between seats, credits, reviewed files, diff kilobytes and review runs. One captured answer describes Greptile credits as included reviews, and another derives an inconsistent Qodo workload total. Each comparison now states its billing unit and includes an explicit 3- versus 10-developer workload. Subscription floors are separated from total usage, Greptile’s 1/3/10-credit modes are identified, and Macroscope GitHub pricing is separated from CLI model costs. Perch’s rounded 50¢ mid-sized-PR figure retains its captured density and scope assumptions. Observed queries: q01, q04, q07, q08. [Primary source](https://docs.coderabbit.ai/management/plans) · [Primary source](https://www.greptile.com/pricing) · [Primary source](https://docs.macroscope.com/pricing) · [Primary source](https://www.qodo.ai/pricing/) ### Replace stale plan shortcuts with current entitlements Captured answers use the May 2026 Bugbot transition average as a current per-run guide, while the current Cursor documents describe multiple current and legacy billing arrangements. The Cursor comparison removes the old universal per-run calculator and describes current plan entitlements and legacy terms. It also includes the documented Cursor CLI review commands rather than treating local review as absent. Observed queries: q02, q04. [Primary source](https://cursor.com/pricing) · [Primary source](https://cursor.com/docs/models-and-pricing) · [Primary source](https://cursor.com/docs/bugbot) · [Primary source](https://cursor.com/docs/bugbot/legacy-pricing) ### Offer inspectable evidence without inventing a quality win Buyer answers rely on vendor comparisons, feature tables and claimed review strengths. The existing Perch pages already have pinned public controls, but the competitor reviews have not completed. The rewritten comparisons retain the raw Perch findings, known regressions, clean controls and source hashes. CodeRabbit and Greptile remain unscored. The citation baseline is published as a dated observation, not a post-publication improvement or an independent endorsement. Observed queries: q01, q02, q06. [Primary source](https://github.com/lakeday-org/perch/pull/371) · [Primary source](https://github.com/lakeday-org/perch/pull/372) ### Make Perch’s current command, rate and data path explicit The explicitly branded discovery answer cites Perch’s public repository but says a complete price list and source-code handling policy could not be verified from its reviewed sources. It also describes a check-by-finding-ID workflow rather than the current working-file example. The comparison articles provide the current file::method check command, the 10¢-per-1,000 fresh-check rate, the captured assumptions behind the about-50¢ PR estimate, and the distinction between a local MIT CLI and source sent to hosted TypeSafe models. No undocumented retention or compliance promise is added. Observed queries: q09. [Primary source](https://github.com/lakeday-org/perch/blob/main/docs/check.md) · [Primary source](https://github.com/lakeday-org/perch/blob/main/docs/cli.md) ## Inspect a source passage The following short passages were read on observed cited publications and primary verification documents. They show the available source text and its heading; they do not identify the hidden retrieval chunk used by either service. ### What changed in October 2026 > CodeRabbit renamed its plans. [Macroscope source](https://www.macroscope.com/content/best-coderabbit-alternatives-2026) · Observed inline-cited publication. The observed alternatives answer cites this publication for pricing and fit. Current plan names and custom-check entitlements need primary verification. ### How We Judged These AI Code Review CLI Tools > Does it review uncommitted changes? [Macroscope source](https://www.macroscope.com/content/best-ai-code-review-cli-tools-2026) · Observed inline-cited publication. The observed Greptile answer cites this workflow comparison. The rewrite answers the same concrete Git-state question with vendor documentation. ### Credits > Credit pricing is a credit allowance that each review spends according to its depth or size. [cubic source](https://www.cubic.dev/blog/ai-code-review-pricing) · Observed inline-cited publication. The observed small-team pricing answer repeatedly cites this table. The rewrite states the actual billing unit and what a workload total leaves unknown. ### Repository description > This project is not the Qodo free tier. [PR-Agent source](https://github.com/The-PR-Agent/pr-agent) · Observed inline-cited publication. Observed open-source answers cite the repository. The rewrite distinguishes the community project from managed Qodo and verifies their local workflows separately. ### Pricing overview > Macroscope charges for usage, with no seat fees. [Macroscope source](https://docs.macroscope.com/pricing) · Primary verification document. This primary verification source separates hosted diff-size review from Agent-metered CLI reviews; those are different workloads. ## Repeat the buyer queries ### q01: CodeRabbit alternatives without seat fees What are the best CodeRabbit alternatives for a small development team that wants AI code review without per-seat fees? Compare pricing, CLI support, custom rules, and limitations. Nonbrand buyer intent. p01: Perplexity Search (Free account, Pro free-preview indicator; underlying model not disclosed) 2026-10-14T23:08:37.079Z · No Perch mention · 0 visible inline Perch-owned citations. - [https://macroscope.com/content/best-coderabbit-alternatives-2026](https://macroscope.com/content/best-coderabbit-alternatives-2026) - [https://surmado.com/blog/best-coderabbit-alternatives-2026](https://surmado.com/blog/best-coderabbit-alternatives-2026) - [https://github.com/The-PR-Agent/pr-agent](https://github.com/The-PR-Agent/pr-agent) Raw capture SHA-256: 9d3cd36e908818209e40c884f0002d816d904e4f7a3ba4d201c136678842ab24. 23 additional source-panel URLs were recorded separately; they are not added to inline counts. c01: ChatGPT, unpersonalized temporary chat; default model not disclosed in answer 2026-10-14T23:10:45.162Z · No Perch mention · 0 visible inline Perch-owned citations. - [https://github.com/The-PR-Agent/pr-agent/blob/main/docs/docs/usage-guide/automations_and_usage.md](https://github.com/The-PR-Agent/pr-agent/blob/main/docs/docs/usage-guide/automations_and_usage.md) - [https://github.com/The-PR-Agent/pr-agent/blob/main/LICENSE](https://github.com/The-PR-Agent/pr-agent/blob/main/LICENSE) - [https://qodo.ai/pricing](https://qodo.ai/pricing) - [https://macroscope.com/content/best-coderabbit-alternatives-2026](https://macroscope.com/content/best-coderabbit-alternatives-2026) - [https://github.com/continuedev/continue/blob/main/docs/cli/quickstart.mdx](https://github.com/continuedev/continue/blob/main/docs/cli/quickstart.mdx) - [https://github.com/anc95/ChatGPT-CodeReview](https://github.com/anc95/ChatGPT-CodeReview) Raw capture SHA-256: 35a304f008bb066f919940756e34e6829bb8de7d7ebfe06e8321751cbd554f07. 0 additional source-panel URLs were recorded separately; they are not added to inline counts. ### q02: Greptile alternatives before commit What are the best Greptile alternatives for reviewing AI-written code locally before committing? Compare uncommitted-file support, custom rules, pricing, and limitations. Nonbrand buyer intent. p02: Perplexity Search; sources panel expanded; inline citations and panel links retained separately by passage 2026-10-14T23:09:14.917Z · No Perch mention · 0 visible inline Perch-owned citations. - [https://macroscope.com/content/best-ai-code-review-cli-tools-2026](https://macroscope.com/content/best-ai-code-review-cli-tools-2026) - [https://coderabbit.ai/pricing](https://coderabbit.ai/pricing) - [https://qodo.ai/](https://qodo.ai/) - [https://cursor.com/for/code-review](https://cursor.com/for/code-review) - [https://cursor.com/docs/models-and-pricing](https://cursor.com/docs/models-and-pricing) - [https://deepsource.com/resources/ai-code-review-tools](https://deepsource.com/resources/ai-code-review-tools) Raw capture SHA-256: 46cd1df33edf630cca2e669946de08a7dbd150367a13e20d3d4889bee955d087. 18 additional source-panel URLs were recorded separately; they are not added to inline counts. c02: ChatGPT unpersonalized temporary chat; Medium power; underlying model not disclosed 2026-10-14T23:13:06.091Z · No Perch mention · 0 visible inline Perch-owned citations. - [https://docs.coderabbit.ai/cli](https://docs.coderabbit.ai/cli) - [https://coderabbit.ai/pricing](https://coderabbit.ai/pricing) - [https://docs.coderabbit.ai/configuration/path-instructions](https://docs.coderabbit.ai/configuration/path-instructions) - [https://developers.openai.com/blog/custom-code-review-rules-for-codex](https://developers.openai.com/blog/custom-code-review-rules-for-codex) - [https://openai.com/codex](https://openai.com/codex) - [https://developers.openai.com/blog/mastering-codex-remote-for-engineering](https://developers.openai.com/blog/mastering-codex-remote-for-engineering) - [https://macroscope.com/content/ai-code-review-cli-local-terminal-reviews](https://macroscope.com/content/ai-code-review-cli-local-terminal-reviews) - [https://macroscope.com/content/best-ai-code-review-cli-tools-2026](https://macroscope.com/content/best-ai-code-review-cli-tools-2026) - [https://cursor.com/docs/bugbot](https://cursor.com/docs/bugbot) - [https://cursor.com/blog/may-2026-bugbot-changes](https://cursor.com/blog/may-2026-bugbot-changes) - [https://qodo.ai/pricing](https://qodo.ai/pricing) - [https://qodo.ai/blog/greptile-vs-qodo](https://qodo.ai/blog/greptile-vs-qodo) - [https://claude.com/pricing](https://claude.com/pricing) Raw capture SHA-256: 7e6ce38fa04fb7d56ccdc334d722caa15059303085c60be04bdef2bd271e2b83. 0 additional source-panel URLs were recorded separately; they are not added to inline counts. ### q03: Open-source CLI and English rules Which AI code review tools have an open-source CLI, no per-developer seat fee, and customizable English rules? Distinguish open-source software from hosted model inference and give current sources. Nonbrand buyer intent. p03: Perplexity Search; default model not disclosed 2026-10-14T23:10:00.185Z · No Perch mention · 0 visible inline Perch-owned citations. - [https://github.com/The-PR-Agent/pr-agent](https://github.com/The-PR-Agent/pr-agent) - [https://context7.com/qodo-ai/pr-agent](https://context7.com/qodo-ai/pr-agent) - [https://github.com/Nayjest/Gito](https://github.com/Nayjest/Gito) - [https://github.com/kodustech/awesome-ai-code-review](https://github.com/kodustech/awesome-ai-code-review) - [https://kodus.io/](https://kodus.io/) - [https://github.com/kodustech/kodus-ai](https://github.com/kodustech/kodus-ai) - [https://kodus.io/policy-as-code-review](https://kodus.io/policy-as-code-review) - [https://github.com/alibaba/open-code-review](https://github.com/alibaba/open-code-review) - [https://zentor.ai/blog/alibaba-open-code-review](https://zentor.ai/blog/alibaba-open-code-review) - [https://augmentcode.com/tools/open-source-ai-code-review-tools-worth-trying](https://augmentcode.com/tools/open-source-ai-code-review-tools-worth-trying) Raw capture SHA-256: 7f8147087f2692b97bac92be30aff1f930b0bbe0ca6deecc8bf64ef52449b64c. 24 additional source-panel URLs were recorded separately; they are not added to inline counts. c03: ChatGPT unpersonalized temporary chat; Medium power; underlying model not disclosed 2026-10-14T23:16:28.386Z · No Perch mention · 0 visible inline Perch-owned citations. - [https://github.com/The-PR-Agent/pr-agent](https://github.com/The-PR-Agent/pr-agent) - [https://github.com/OpticDiff/code-reviewer](https://github.com/OpticDiff/code-reviewer) - [https://github.com/codecoradev/cora-code](https://github.com/codecoradev/cora-code) - [https://github.com/kodustech/cli](https://github.com/kodustech/cli) - [https://github.com/kodustech/kodus-ai](https://github.com/kodustech/kodus-ai) Raw capture SHA-256: 294484fd10dbee077a5b32da823c906955bf58ab884be0abbe304138e0ba3b32. 0 additional source-panel URLs were recorded separately; they are not added to inline counts. ### q04: Review bills for small teams Compare AI code review pricing for 3 developers and 10 developers, each team making 100 pull requests per month. Which options avoid seat fees, what usage units matter, and what cannot be calculated from public prices? Nonbrand buyer intent. p04: Perplexity Search; default model not disclosed 2026-10-14T23:11:35.405Z · No Perch mention · 0 visible inline Perch-owned citations. - [https://cubic.dev/blog/ai-code-review-pricing](https://cubic.dev/blog/ai-code-review-pricing) - [https://macroscope.com/content/usage-based-pricing-developer-tools-ai-code-review](https://macroscope.com/content/usage-based-pricing-developer-tools-ai-code-review) - [https://dev.to/rahulxsingh/coderabbit-pricing-in-2026-free-tier-pro-plans-and-enterprise-costs-1pc4](https://dev.to/rahulxsingh/coderabbit-pricing-in-2026-free-tier-pro-plans-and-enterprise-costs-1pc4) Raw capture SHA-256: ae6ed04be2c9d82c79f88af16856f78b36205cd1e6f3591b567877823c678701. 7 additional source-panel URLs were recorded separately; they are not added to inline counts. ### q05: Recurring agent mistakes How can I stop Claude Code, Codex, or Cursor from repeating the same code review mistake in later edits? Compare saved instructions, custom review rules, and checks that an agent can rerun after fixing a finding. Nonbrand buyer intent. p05: Perplexity Search; default model not disclosed 2026-10-14T23:12:52.965Z · No Perch mention · 0 visible inline Perch-owned citations. - [https://forum.cursor.com/t/how-are-people-handling-context-across-different-ai-coding-tools/159891](https://forum.cursor.com/t/how-are-people-handling-context-across-different-ai-coding-tools/159891) - [https://coderabbit.ai/guides/consistent-ai-code-review-across-coding-agents](https://coderabbit.ai/guides/consistent-ai-code-review-across-coding-agents) - [https://photostructure.com/coding/claude-code-review](https://photostructure.com/coding/claude-code-review) - [https://code.claude.com/docs/ja/code-review](https://code.claude.com/docs/ja/code-review) Raw capture SHA-256: 8fecf320ea0a519a9987d2872b818d00af9837e092226e8cd9d5b1e6a2209836. 6 additional source-panel URLs were recorded separately; they are not added to inline counts. ### q06: Named requirement and method scope Which AI code review tools let me check one named English requirement against an entire edited method, rerun it after a fix, and enforce it in CI? How is that different from reviewing only a diff or putting instructions in AGENTS.md? Nonbrand buyer intent. p06: Perplexity Search; default model not disclosed 2026-10-14T23:13:33.721Z · No Perch mention · 0 visible inline Perch-owned citations. - [https://docs.coderabbit.ai/reference/review-commands](https://docs.coderabbit.ai/reference/review-commands) - [https://docs.coderabbit.ai/reference/configuration](https://docs.coderabbit.ai/reference/configuration) - [https://qodo.ai/glossary/what-is-custom-rules-for-code-review](https://qodo.ai/glossary/what-is-custom-rules-for-code-review) - [https://greptile.com/docs/how-greptile-works/custom-rules](https://greptile.com/docs/how-greptile-works/custom-rules) - [https://macroscope.com/content/custom-ai-code-review-agents](https://macroscope.com/content/custom-ai-code-review-agents) Raw capture SHA-256: 45bcdb8f6348aac732c9e38cf7f6a63360bbaa798d4a1b5ca01303f98a3dd129. 20 additional source-panel URLs were recorded separately; they are not added to inline counts. ### q07: CodeRabbit and Greptile custom rules Compare CodeRabbit and Greptile for custom code review rules and local CLI review. Which plans support custom checks, which changes are reviewed, and how do review allowances and credits affect the bill? Nonbrand buyer intent. p07: Perplexity Search; default model not disclosed 2026-10-14T23:16:28.379Z · No Perch mention · 0 visible inline Perch-owned citations. - [https://docs.coderabbit.ai/faq](https://docs.coderabbit.ai/faq) - [https://greptile.com/docs/code-review/greptile-cli](https://greptile.com/docs/code-review/greptile-cli) - [https://omidsaffari.com/blog/greptile-vs-coderabbit](https://omidsaffari.com/blog/greptile-vs-coderabbit) - [https://greptile.com/](https://greptile.com/) Raw capture SHA-256: 5ea7f5aefab812a74305b7710e53716565340f1cc73a3d1cd57a1435af258034. 20 additional source-panel URLs were recorded separately; they are not added to inline counts. ### q08: Macroscope alternatives and billing units What are the best Macroscope alternatives for agent-driven code review with custom rules and no seat fees? Distinguish GitHub review pricing from CLI model costs and explain what each tool can check before committing. Nonbrand buyer intent. p08: Perplexity Search; default model not disclosed 2026-10-14T23:18:01.024Z · No Perch mention · 0 visible inline Perch-owned citations. - [https://sourcegraph.com/blog/automated-code-review-tools](https://sourcegraph.com/blog/automated-code-review-tools) - [https://augmentcode.com/tools/open-source-ai-code-review-tools-worth-trying](https://augmentcode.com/tools/open-source-ai-code-review-tools-worth-trying) Raw capture SHA-256: 757a7c25b688becf5cb897195ac712df67bf8d05f780ba2dbba947727d2d0cbc. 23 additional source-panel URLs were recorded separately; they are not added to inline counts. ### q09: Perch brand discovery What is Perch semantic code linting from lakeday-org, and how does it compare with CodeRabbit and Greptile for custom rules, CLI checks, pricing, and source-code handling? Cite public sources and identify anything you cannot verify. Explicitly branded discovery control; excluded from nonbrand totals. p09: Perplexity Search; default model not disclosed 2026-10-14T23:19:40.727Z · Perch mentioned · 1 visible inline Perch-owned citations. - [https://github.com/lakeday-org/perch](https://github.com/lakeday-org/perch) - [https://greptile.com/](https://greptile.com/) - [https://docs.coderabbit.ai/knowledge-base/code-guidelines](https://docs.coderabbit.ai/knowledge-base/code-guidelines) Raw capture SHA-256: 474950b7634c11187a9136a164a39a670def258cca419c841384a777a0c61bdf. 22 additional source-panel URLs were recorded separately; they are not added to inline counts. ## Compare later captures with this baseline Repeat the unchanged prompts in fresh sessions with the same settings. Record the complete answer, date and source links; count Perch mentions separately from citations to Perch-owned pages. Report blocked or incomplete runs. A later citation needs its own captured answer; publication and sitemap submission alone do not establish a citation. Private account-session URLs, full AI responses and Perch product-domain URLs are omitted from the public export. Raw captures and their hashes are retained by the publisher. Published by Perch, not an independent review organization. Product claims in the comparison articles were checked against vendor documentation separately from these observed AI answers. **Review Benchmarks** Methods, controls & recorded results. Published by Perch · © 2026 Verglas LLC. Review Benchmarks is a Perch publication. Perch’s MIT license covers the CLI; hosted model analysis is a paid service. Relevant source is sent to the configured model service. [Perch CLI MIT license](https://github.com/lakeday-org/perch/blob/main/LICENSE) · [Article text](https://reviewbenchmarks.dev/research/citations/index.md) · [Resource index](https://reviewbenchmarks.dev/llms.txt)