Response quality (Cipher)

Cipher settings and plans

Choose how deeply Cipher checks a survey, switch individual checks on or off, and see what each plan includes.

Cipher is on for every new survey, at the Pro level by default. On the Free plan it runs at Basic. You can change the level, switch off individual checks, or turn Cipher off for a survey.

Open the Cipher settings

  1. Open your survey and go to Settings. From the editor, open Quick settings and click All project settings.
  2. Choose the Cipher tab.
  3. Make your changes and click Save changes.

Settings apply to responses submitted after you save.

Turn Cipher on or off

Under Enable Cipher, the Check response quality on this survey box controls whether Cipher scores this survey's responses.

When Cipher is off, responses are not scored. Responses can still carry flags from the invisible bot check and from any integrity settings you switched on (see below).

Choose an intelligence level

Under Intelligence level, pick one of three levels. Each level includes everything in the level below it. Every flag comes with plain-language reasoning at every level.

LevelWhat it addsPlan
BasicSpeed and timing, straight-lining and careless patterns, attention checks, hidden trap fields, pasting, bots and automation, and device spoofing.Every plan, including Free
ProHow each written answer was produced (typed, revised, pasted or copied out, how fast and when), hidden characters, typing rhythm, tab switching and leaving the page, pointer, hover and scroll behaviour, sentence rhythm, contradictions between answers, copied text, and a light AI review of answer quality and meaning.Hive
MaxChecks across responses (fraud rings, shared answers, coordinated timing, shared devices, reputation across your surveys), network checks (VPN, datacenter, Tor, proxy, timezone), a thorough AI review, and a web search for long written answers that match text already published online.Swarm and Enterprise

The Max web search reads the question first. Questions that ask people to quote, cite sources, or paste text are skipped.

Levels your plan doesn't include show a lock. If a survey is set to a level above your plan, it runs at the highest level your plan allows, and the settings page tells you which one, for example "Your plan runs this survey at Basic. Upgrade to use Pro."

How Cipher spots AI-written answers

Cipher catches AI-written answers from how people behave. It looks at how each answer originated and where it came from, whether it was typed or pasted, how quickly it arrived, and when. The AI review reads the wording on top of that.

Credits

Basic is free and shows Free in the level list. Pro and Max cost credits for each response Cipher checks, shown in the level list: 0.3 a response at Pro and 6 at Max. Max costs more because it runs the thorough AI review and the web search.

The cost comes out of your plan's Cipher credits first (500 a month on Hive, 2,000 on Swarm), then your general credits. If both run out, responses are still collected and checked at the free Basic level until your credits refill. See Credits.

What each plan includes

  • Free: Cipher Basic on every response, including speed, careless answer, attention, pasting, bot and device checks, and the invisible bot check.
  • Hive: Cipher Pro, which adds how each written answer was produced, AI-written answer detection, typing and attention behaviour, and contradictions.
  • Swarm: Cipher Max, which adds checks across responses such as fraud rings, shared devices, shared answers and coordinated timing, VPN, proxy and Tor detection, and the web search for copied answers.
  • Enterprise: everything in Max, with custom credits.

Switch individual checks on or off

Under Individual signals, you can see every check, grouped into Behavioral, Device, Content, Network and AI signals, with a count of how many are active.

  • Untick a check to stop it running on this survey.
  • Use Select all and Deselect all for everything at once, or the buttons next to each group for that group.
  • Checks that your level doesn't include are greyed out and say which level they run from, for example "Runs from Max."

You can only switch checks off within your level. Changing the level resets your individual choices to that level's defaults.

Tell Cipher what a question expects

Cipher reads each question and the text around it to decide whether respondents should leave the page or paste their answer there. You can set this yourself when it gets a question wrong.

Select the question in the editor, open the Validation tab and find Cipher:

  • Leaving the page. Choose Expected when the question sends people to another site or app. Time away on that question won't count against them.
  • Pasting. Choose Expected when people should paste their answer, such as a link or a prompt. The answer is still checked for AI-written text.

Auto lets Cipher decide from the wording. Not expected tells Cipher to judge the question normally even if its wording suggests otherwise.

You can also ask Surbee's AI in the editor's chat, for example "People will leave the page on question 4 to try the app" or "Question 6 asks them to paste a link."

This setting only affects the question it is on. Leaving or pasting elsewhere in the survey is still checked.

Collect behavioral data

In the editor, open Quick settings and the Responses tab. The Collect behavioral data switch (time per question, mouse trails, fraud signals) is on by default.

When you switch it off:

  • Surbee stops recording how the form is used in the respondent's browser: timing, mouse, keystroke rhythm and attention.
  • Surbee doesn't keep data on people who leave without finishing, so Where people left in the Funnel view has nothing to show.
  • Respondents don't see the integrity notice at the bottom of the form.
  • Cipher has far less to work with, so its scores are much less informative.

Integrity settings

The Integrity tab in Quick settings has extra controls that sit alongside Cipher:

  • Randomize answer options.
  • Flag copy and paste. Pasting is allowed but recorded on the response as a flag. The source site cannot be captured.
  • Block VPNs and proxies. Refuses submissions from known VPN, proxy, Tor and datacenter connections. If the lookup is unavailable, everyone is let through.
  • Flag IP changes. Flags responses whose network changes mid-session.
  • Detect collusion. Flags respondents with similar answers and synchronised timing.
  • Network restriction. Limits responses to the networks you list, one IP address or range per line, for example a campus network.

Was this article helpful?