01
Technical search access
HTTP status, title, description, canonical, indexing directives, H1, language, hreflang, JSON-LD block count, a sitemap sample and crawler rules. One public URL per monitor.
Starry Monitor
Know when search and AI answers change.
Starry Monitor checks the page you publish and the answers AI engines give about your category. It keeps the evidence, shows what moved, and tells you which change to make next.
- AI engines supported
- 6
- Default run
- Weekly
- Kept with every answer
- Evidence
- Starting plan
- $29
The same prompts, on a schedule
Pasting one answer into a checker tells you about that one answer. Monitor sends the same prompt set to the same engines every week, or every day, and stores each answer. A change shows up as a difference between two runs you can open side by side.
Each run records
- whether the brand or one of its spellings was named
- the sentence in the answer where it first appears
- whether the answer cited a URL on your site
- which listed competitors were named, and in which sentence
- for the website check, status, canonical, indexing rules and crawler access
What it monitors
Four observations that answer different questions. A technical check says whether the page can be read. An AI run says what an engine actually answered. The change list says what moved since last time.
02
AI answers
Run your prompt set across connected engines. Store the answer, the brand mention, where it appears, the cited URLs and the competitor names in the same answer.
03
Change alerts
A new noindex, a changed canonical, a blocked crawler, a lost mention or a dropped citation becomes an alert with the before and after.
04
A report you can check
Every result keeps the source URL, engine, prompt and time. Open the evidence and compare it with the previous run. No invented visibility score.
How a monitor works
01
Add the page and the brand
Use one public URL and the name people would see in an answer. Add spelling variants and up to ten competitors.
02
Add the questions buyers ask
Five to twenty prompts are enough to see a pattern. Use the words a buyer would type, not a keyword with “how do I” added.
03
Run it and read the evidence
A weekly run collects answers and technical changes. Run on demand when you publish a fix or a new page.
04
Fix one thing, then re-run
Start with a blocked crawler, a broken canonical or a missing description. The next run shows whether the observation changed.
What it cannot promise
No tool can make an AI engine cite a page or keep a position. Monitor records what an engine answered and what changed since the last run. A sentence number counts sentences inside one answer and is not a ranking.
What the free trial includes
One monitor, five prompts, two engines, four runs a month, 100 AI answers and four website checks. No card is needed. When an allowance runs out, runs pause until the next calendar month (UTC) or until you move to a paid plan.
Plans
The trial is free. Paid plans raise the number of monitors, prompts, engines and monthly answers. A run sends every active prompt to every connected engine, and each answer counts toward the monthly answer limit.
Trial
Free
One site and one small prompt set.
- 1 monitor
- 5 prompts
- 2 engines
- 4 runs a month
- 100 AI answers
- 4 website checks
Starter
$29a month
A founder checking one product and a few competitors.
- 3 monitors
- 10 prompts each
- 4 engines
- 20 runs a month
- 600 AI answers
- 30 website checks
Pro
$79a month
A small team watching several pages and markets.
- 10 monitors
- 20 prompts each
- 6 engines
- 60 runs a month
- 5,000 AI answers
- 100 website checks
Agency
$199a month
Client work with separate evidence for each site.
- 25 monitors
- 20 prompts each
- 6 engines
- 150 runs a month
- 12,000 AI answers
- 300 website checks
Prices are in US dollars a month. Choosing a plan opens a confirmation in the workspace first. Payment happens on Stripe only after you confirm there.
What it does not do
- It does not track Google rankings. It records what an AI answer said and which sources it cited.
- AI answers change between runs. Read a share across prompts and runs, not one answer as a rank.
- A mention is not a citation. A citation is not a click. Each is stored separately.
- DeepSeek and Kimi answers without live web search are marked ungrounded.
- One monitor checks one public URL. Add another monitor for another page or site.
- The technical check reads the initial HTML and public robots.txt. It does not prove indexing or ranking.
Questions before you start
Does Starry Monitor guarantee that ChatGPT will cite my site?
No. It records whether a connected engine named or cited you, and what changed. It cannot make an engine cite a page.
How is this different from a rank tracker?
A rank tracker records a position for a keyword. Starry Monitor records the answer, the brand mention, the cited URLs and the competitor names for a prompt.
Which AI engines can I monitor?
OpenAI, Perplexity, Gemini, Anthropic, DeepSeek and Kimi. The workspace shows which engines are connected. An engine without a server key is not run.
How often does it run?
Weekly by default. Daily is available while the plan's monthly run and answer allowance lasts.
What does the technical check inspect?
HTTP status, title, meta description, canonical, robots directives, H1 count, language, hreflang, JSON-LD block count, a sitemap sample and crawler rules for Googlebot, OAI-SearchBot, PerplexityBot and Claude-SearchBot.
What happens when an engine fails?
The run is marked partial and the error is stored with the engine and prompt. The other engines still finish.
Can I use it for client work?
Agency has 25 monitors, so each client site keeps its own prompts and evidence. The API returns the same records the workspace shows.
Is my monitor data public?
No. Monitor data is tied to your account. Reports are not indexed and are not shown on the public site.