Judge a Sydekick answer
How do I tell whether a Sydekick answer is trustworthy?
Sydekick shows you its work, not its certainty. It does not put a confidence score beside an answer or a recommendation. What you get instead is the scope and skill each of your messages ran under, a live activity summary while it works, and, on some answers, the source it read. Those are what you check before you act on a reply.
Key facts
| Confidence score | None. No percentage and no low-confidence warning on replies |
| Private reasoning | Not shown. Sydekick does not show a transcript of how it decided |
| Under your message | The scope you asked under and the skill you pinned, with the time. Kept with the conversation |
| While it works | Plain-language status lines such as Thinking and Checking listings & reviews |
| After a live reply | A collapsed Worked 24s · 2 steps row. Not kept when you reopen the conversation |
| Some answers | Attach what they read, such as a review with a View on Google link |
| The durable record | Agent → Audit logs |
Why this matters
A fluent answer can still be about the wrong client, based on the wrong page, or out of date, and a score would not catch any of that. Ask what Sydekick was asked, what it looked at, and whether the number matches the screen it came from. Synup shows you the first two; the third is your thirty-second check.
What Sydekick actually shows you
Four things, in the order you meet them.
The scope under your own message. Each message you send carries a line beneath it with the client and location you scoped it to, the skill you pinned if any, and the time it was sent. It appears when the conversation was scoped to a client or location, and it stays with the conversation, so you can still read it a week later. It is on your message, not on Sydekick's reply: it records what you asked under, which is what to check first when an answer feels like it belongs to someone else. See choosing the client, location and skill.
Live status lines. While it works, Sydekick narrates in plain words such as Thinking, Working on it, Checking listings & reviews, Reading your reviews, Pulling SEO data, Searching the web, Building report, Generating image and Sending for your approval. These show what Sydekick is doing; they are not a quality or completeness score.
A work row on a live reply. When the reply arrives, the narration collapses into one row above it, such as Worked 24s · 2 steps, which opens to list the steps. It tells you how long the task took and how many distinct activity lines were recorded. A higher count means more recorded steps, not more sources or a more reliable answer.
The things it attached. Some answers carry what they were built from, and these are kept with the reply:
- A review it discusses appears as a card with a View on Google (or the review's own site) link and See all reviews → to the full list.
- An AI visibility or map-ranking answer can quote the answer engine's own reply or show the results grid, and says Not mentioned or The answer wasn't captured for this report. when a source had nothing.
- Files it saved stay as file cards. Work waiting for your decision shows as an approval card.
When Sydekick asks first
Sometimes Sydekick pauses to ask. A card reading A couple of quick questions offers answers to pick from. Skip — just build it carries on without them, and the card says Without answers, Sydekick will use defaults and make assumptions. If you skip, check the assumption it made. After a failed turn, Ask again sends the same request once more.
Some pages also offer a Show me walkthrough that opens the page where the setting lives and points to it, so you can see the answer in the product rather than taking it from the chat.
What it does not show
A confidence score. Sydekick does not label an answer or recommendation with a percentage.
Its private reasoning. The work row summarises the activity, not a transcript of how Sydekick chose between interpretations.
A permanent copy of the work row. The live Worked … · … steps row is shown when the reply arrives. If you reload the page or reopen the conversation from the list, your scope line, the reply and the cards stay, but the row does not come back. If you need the activity afterwards, read Agent → Audit logs, described in reading the Agent audit log. Long jobs can also post progress notes into the conversation as ordinary messages, which stay.
The thirty-second check
For anything you will act on or send to a client:
Read the scope line
The line under your message shows the client and location it ran under. If it is not the one you meant, scope the conversation and ask again rather than mentally adjusting the answer.
Read the work
While it runs, the status lines show what it is doing. On a live reply the Worked row lists the steps. If the activity does not match the task, check the answer where the information lives.
Open what it points to
A review card links to the review. A file opens in Canvas. Open it and read the source rather than the summary.
Confirm the number on its own screen
Credits are under Agent, post results under Social, directory counts under Listings, ratings under Reviews. Open the screen that owns the figure and check it matches.
Errors that matter for trust
If Sydekick cannot finish, it says so rather than guessing: Something went wrong with the reason, Your agent stopped responding. Retry?, or You're out of credits. Top up to keep chatting. Treat a partial answer after one of these as unfinished.
Watch out for
- A confident tone is not evidence. Sydekick writes the same way whether the data was complete or not.
- The scope line is on your message. Look under your own message, not under the reply, to see what it ran under.
- No work row after a reload is normal. It is only shown with a live reply.
Questions people ask
Where is the confidence score I was expecting?
There isn't one on a Sydekick reply. Read the scope line under your message, the work it shows, and the source it attaches instead.
The step count disappeared from an old conversation. Is that a bug?
No. The row is shown with the live reply and is not kept when you reload or reopen the conversation. The reply, your scope line and any cards stay. For a durable record of what the agent did, read Agent → Audit logs.
Why is there no scope line under my message?
It only appears when the conversation was scoped to a client or location. An unscoped conversation runs across All clients.
Can I ask Sydekick how sure it is?
You can, but it is not a reliable check: a model asked to vouch for its own answer will tend to agree with itself. Check the source and the screen instead.