FAQ

How DevTune measures what AI says about you, what to fix, and whether it worked.

Product

What is DevTune?

DevTune changes what AI platforms tell your buyers about you. It runs one loop: measure what ChatGPT, Perplexity, Gemini, Google AI Mode, and Microsoft Bing Copilot say and cite, see what those systems fetch from your site and who they send, rank the fixes, then prove what moved after you ship them.

Who is DevTune for?

Companies that care what AI says about their product. We focus on developer-tool and infrastructure companies. The same loop works for any company whose buyers now ask an AI before they ask a search engine.

How is DevTune different from SEO tools?

SEO tools rank pages in Google. DevTune records the answers AI platforms actually give, the URLs they cite, the pages they fetch, and the visits they send. Traditional search still matters, which is why Search Console and GA4 sit beside the AI numbers. They are not the same channel.

What does measure, fix, prove mean?

Measure is Visibility and AI Demand, from scheduled probes and first-party request data. Fix is Act, the ranked queue of actions, the answer-engine readiness audit, and drafts that wait for a human. Prove is Outcomes, the before/after window on citations, fetches, and referrals for every shipped action. Every result keeps its source coverage and comparison dates visible.

Does DevTune track GitHub stars and package downloads?

When you add GitHub and registry URLs to a project, DevTune still records stars and downloads as supporting signals. They are not a product pillar. The loop is visibility, AI Demand, act, and outcomes.

Visibility

Learn more

What does Visibility track?

Scheduled probes ask the questions your buyers ask on ChatGPT, Perplexity, Gemini, Google AI Mode, and Microsoft Bing Copilot. DevTune records the answer, the citations, brand mentions, and which competitor was cited instead. Community Discourse watches related threads on Dev.to, Stack Overflow, Hacker News, GitHub, and Reddit.

Which AI search platforms does DevTune probe?

ChatGPT, Perplexity, Gemini, Google AI Mode, and Microsoft Bing Copilot. Starter and Plus include ChatGPT, Perplexity, and Gemini. Pro and Agency/Enterprise add Google AI Mode and Microsoft Bing Copilot.

How do prompts work?

A prompt is a tracked query, such as "What's the best authentication library for Next.js?". You pick the set during project setup and can add more later. DevTune runs those prompts daily across the platforms on your plan. Starter includes 25 prompts, Plus 50, Pro 100, and Agency/Enterprise 300 per project.

What's the difference between a citation and a mention?

A citation is a URL in the answer. A mention is your brand named without a link. They are counted separately. Presence rate uses citations of your tracked URLs. Mentions do not count as presence.

What is share of voice?

Share of voice is your primary citations divided by your primary citations plus the citations of the competitors you configured, over a prompt set and a window. Citations of other sources are tracked, but they are not in that denominator. This is your share against the rivals you chose, not your share of everything cited.

How does competitor tracking work?

Each project has a primary brand and competitor brands, each with tracked URLs and brand terms. The same prompts run for all of them, so you can see which prompts they win, which pages get cited, and how that shifts over time.

What is Community Discourse?

Threads your buyers already read, on Dev.to, Stack Overflow, Hacker News, GitHub, and Reddit. DevTune surfaces discussions that overlap your tracked URLs or AI citations, and clusters where you and a competitor appear together. Reddit comment-level coverage is on Agency/Enterprise.

AI Demand

Learn more

What is AI Demand?

What AI systems do with your site: training crawls, live answer fetches, and the referral sessions that follow. It also records 404 demand, pages those systems keep requesting that do not exist.

What's the difference between a crawl, an answer fetch, and a referral?

A crawl is indexing. An answer fetch is an AI system retrieving a page to build an answer. A referral is a human or agent arriving from an answer. DevTune counts all three separately.

What is 404 demand?

AI systems repeatedly request URLs you never published. Those misses are a content-creation signal from real requests, not from a keyword tool. Only a server-side sensor sees them. The browser snippet cannot.

How do I collect AI Demand data?

Three sensors, and they are not interchangeable. A Cloudflare log pull covers hosts already proxied through Cloudflare. Edge middleware covers Vercel and other app hosts. A browser snippet catches referrals and whichever bots run JavaScript.

Only Cloudflare and edge middleware see server-side answer fetches and 404 demand. Connect one crawler sensor per hostname, plus GA4 for conversions.

Does exceeding the AI Demand allowance stop measurement?

No. The monthly event allowance covers raw event detail. If you go over it, rollup-backed counts, trends, and Outcomes keep working. What pauses is raw event retention and 404-demand classification for the rest of the period.

Event history is how far back raw events stay queryable: 30 days on Starter, 90 on Plus, 180 on Pro, and 365 on Agency/Enterprise.

What is Act?

A ranked queue of fixes, each carrying the evidence that put it there: lost citations, uncited demand, weak pages. An answer-engine readiness audit feeds the queue. Actions can carry a ready-to-edit brief. Nothing publishes itself.

What does the audit score?

Crawlability, extractability, llms.txt, and citability. It is built for answer engines, not only Google. Findings are re-checked and wired into the action queue.

What are locked pages?

Pages you mark off-limits. Every suggestion steers around them and says that it did. Use this for pages legal or brand need left alone.

Do I have to take DevTune's suggestions?

No. You can snooze, dismiss, adopt into the backlog, or create your own actions. Agents can draft the work. A person still decides.

Outcomes

Learn more

What is Outcomes?

Every shipped action gets a before/after window on citations, answer fetches, and referrals. The ledger reports what changed and shows the source coverage behind each result.

How does DevTune handle incomplete data?

DevTune shows which source or part of the comparison window is missing. It does not turn missing coverage into a zero or fill the gap with an estimate.

How long is the measurement window?

The comparison uses the 14 days after a detected change against the 14 days before it. Open windows stay labeled as measuring until enough time has passed.

What can agents do?

Investigate a citations drop, audit a page, rank what to refresh, draft a rewrite, watch a competitor, or send a weekly report. They run on demand or on a schedule, in the app or from Slack.

Do agents publish to my site?

No. Every draft waits for a human to approve it. Agents do not publish, deploy, or change your live content on their own.

How do agent credits work?

Credits power agent runs and brief generation. Each plan adds a monthly allowance: 10 on Starter, 20 on Plus, 50 on Pro, and 100 per project on Agency/Enterprise. The allowance expires at the next billing anniversary. Bought packs never expire. Spending uses the expiring allowance first.

Can I run agents from Slack?

Connect Slack from notification channels, then invoke an agent, follow the run in the thread, and chat with it there. Every draft still waits for approval. Slack is on every paid plan.

API and MCP

Learn more

Can my own agents use DevTune?

Plus and above include a scoped REST API and an MCP server. Your automations and agents read the same ground truth as the app: visibility, citations, AI Demand, actions, and outcomes. API keys are project-scoped.

What's the difference between the API and MCP?

The REST API is for programmatic reads, webhooks, and your own tooling. The MCP server is for agents in tools such as Claude, Cursor, and Codex. MCP supports OAuth for interactive clients and API keys for non-interactive ones.

Billing and plans

Learn more

What plans do you offer?

Starter is $49/mo ($499/year) with a 7-day trial. Plus is $99/mo ($999/year). Pro is $199/mo ($1,999/year). Agency/Enterprise is $399 per project per month ($3,999/year). Every paid plan includes Visibility, AI Demand, Act, and Outcomes. Plans differ by how much you measure and how many agent credits you get.

What's included on every plan?

Scheduled probes, citations, share of voice, competitor gaps, Community Discourse, AI Demand, the readiness audit, ranked actions with drafts, Outcomes, agents in the app and in Slack, and alerts.

Is there a free trial?

A 7-day trial on Starter monthly. A card is required. You pay $0 today and can cancel before it ends.

What happens if I reach plan limits?

Prompt and project limits cap how much you can configure. Hitting the AI Demand event allowance does not stop rollup-backed measurement. Agent credits can be topped up with one-time packs.

How often is data updated?

AI search probes run daily on every plan. Crawler and snippet events arrive as they happen. Search Console and GA4 sync on a schedule and fill in history afterwards.

Can I cancel my subscription?

Cancel from Billing in account settings. You keep access until the current period ends, and you can reactivate later.

Where can I find my invoices?

Account settings, under Billing. Invoices are generated automatically and can be downloaded as PDFs.

What payment methods do you accept?

Visa, Mastercard, and American Express. Annual Agency/Enterprise plans can invoice.

If you have any questions, please contact us