How to use Snacked

Snacked turns a one-sentence brief into a finished quiz, test, survey or two-axis compass — questions, outcomes, scoring and result copy. This guide walks you from signup to your first published piece, and then through every feature you'll grow into.

If you prefer to skim, jump to a section below. If you'd rather feed this to your favourite LLM, the same content is available at /guide.md, and a machine-readable index lives at /llms.txt.

Table of contents

• What Snacked is, in one minute
• Sign up and the dashboard
• Create your first quiz with AI
• Edit anything the AI produced
• The scoring matrix, demystified
• Two ways to iterate
• Theme and brand your quiz
• Publish and share
• Turn a result into an invite
• Collect emails
• Connect your email tool
• Add branching logic
• Embed your quiz anywhere
• Track conversions
• Read and export responses
• Plans, limits and billing
• Languages
• Troubleshooting
• What's next

What Snacked is, in one minute

Snacked is an AI maker for four kinds of interactive content:

• Personality quiz — sorts people into types they recognise themselves in.
• Graded test — scores answers against a key, with pass and fail bands.
• Survey — collects answers with no scoring; you read the aggregate.
• 2-axis compass — places people on two axes and lands them in a quadrant.

You pick the format and describe in one sentence what it should do. Snacked generates the questions, the options, the outcomes and every scoring weight in a single pass. You edit anything, publish to a shareable URL, and read every response in your analytics.

Responses are unlimited on every plan and the analytics are complete on every plan. The email gate runs on every plan too: a response that gives an email is a lead, and free accounts collect their first 25 in the clear. Past 25 the gate keeps capturing — the new leads land locked, with the address hidden — and upgrading unlocks every one of them at once. Pro is the rest of the funnel: 10,000 leads a month, each delivered to your email tool, result CTAs, and the removal of the "Made with Snacked" badge.

Snacked is intentionally niche-agnostic. Fitness coaching, B2B software, a university course, an internal team survey — the same builder works.

Sign up and the dashboard

Sign up at snacked.it with email or a third-party identity provider. No credit card required — the Free plan stays free for five published quizzes, with unlimited responses.

After signup you land on your dashboard. From here you can:

• Create a new quiz (button top-right)
• Open any existing quiz to edit, view responses or change settings
• Open Settings to manage your account, billing, branding defaults and integrations

If you ever lose your way, the Snacked logo in the header always takes you back to the dashboard.

Create your first quiz with AI

Click New quiz. You choose between generating with AI and starting from a blank builder. Both start by picking the format — quiz, test, survey or compass — because the format decides what gets built and can't be switched afterwards.

For the AI route you fill in:

1. What it should do. One or two sentences. Specific beats clever. Example: "Tell solo founders whether to hire a marketer or do it themselves."
2. Context (optional). Anything that steers the copy: the subject, the audience, your angle.
3. How many result types (personality only). Four to eight outcomes.
4. Tone and language. Snacked respects the language end-to-end — questions, outcomes, result copy and button labels.

Hit Generate. The AI streams it live: title, questions, options, outcomes, the full scoring matrix and result copy. You watch it appear in real time.

You don't set the length — it's read off what you asked for. Say it's a quick thing for a newsletter and you get a short quiz; describe an assessment for clients and you get a longer one. Every extra question is another chance for someone to close the tab, so with no signal either way it errs short. The outer bounds are 6–12 questions for a personality quiz, 5–15 for a test, 6–15 for a compass and 3–15 for a survey — and in the editor you can add or remove as many as you like afterwards, with no limit at all.

Starting from blank gives you the same skeleton with nothing filled in — a first question, empty outcomes for the format you picked — and you write it yourself.

Most generations are usable as-is. But everything is editable — keep reading.

Edit anything the AI produced

Open any quiz from your dashboard. The editor shows the tabs that make sense for its format — a survey has no scoring, a test has score bands, a compass has axes and regions:

Content

The question list. For each question you can edit the text, change its type (single choice, multiple choice, free text or rating), edit each option's text, add or remove options, drag to reorder, or delete the question. On a graded test you also mark the correct option here. Add a question with the + Add question button. Snacked supports 3 to 12 questions — fewer feels rushed, more loses people.

Outcomes

What a response resolves to. The tab is named for your format:

• Personality → Archetypes. Four to eight result types, each with a title, markdown result copy and an optional CTA URL. Adding or removing an archetype also adds or removes its scoring dimension.
• Test → Bands. Score ranges with a pass mark: a name, a minimum and maximum score, and whether the band counts as a pass.
• Compass → Axes and Regions. The two axis labels, and one region per quadrant with its own copy.
• Survey → nothing. There is no outcome to configure; respondents get a thank-you and you read the aggregate.

Scoring

The scoring matrix — the heart of anything that resolves to an outcome. See the next section. A survey has no Scoring tab; a compass shows two, one per axis.

Theme

Per-quiz colours, fonts and presets. Covered in Theme and brand your quiz below.

Settings

Per-quiz settings: title, description, social preview image, the intro screen, the email gate, integrations, branching rules, tracking pixels, embed code.

The intro screen is optional. By default a quiz opens on a screen with its title and a Start button. Turn it off and people land straight on the first question — answering it is what starts the quiz. It's worth trying: on our own best-performing quiz, 41% of visitors pressed Start and 91% of those finished, which means the intro was losing more people than every question combined. It asks for a click and gives nothing back. Keep it when the framing matters — an assessment whose answers depend on people understanding what they're being asked, or a quiz someone arrives at cold with no idea what it is.

The scoring matrix, demystified

Every answer contributes to every outcome. If someone picks option B on question 3, that single choice nudges their score on all the archetypes, not just one.

Concretely: each option carries one numeric weight per scoring dimension. At the end, Snacked sums the weights for the picks that were made and resolves a result:

• Personality — the highest-scoring archetype wins.
• Test — the total is compared against your score bands (weight 1 for the correct option, 0 for the rest).
• Compass — the x and y totals place the respondent in a quadrant.
• Survey — nothing is scored.

The hard rule for generation: every option must have a weight for every dimension. No gaps, no zeros-by-default. If the AI returns a sparse matrix, the generation fails and you regenerate — by design, to keep results meaningful.

You don't need to touch the matrix to ship — but when you do, there are two ways in.

Under every answer in the Questions tab there is a Points to picker. Choose an outcome and that answer counts fully towards it and not at all towards the others. That is what most quizzes want, and it is one click per answer instead of one number per outcome. On a compass you get two small −2…+2 pickers instead, one per axis.

The Scoring tab is the full grid, for when an answer should count towards more than one outcome at once. Both write the same weights.

Questions also warns you when an outcome has nothing pointing at it — no respondent can ever land on an outcome with no weights, and that is easy to create by accident and hard to notice.

Two ways to iterate

Try again, with notes. Every quiz has a Try again button next to Preview. Say what isn't working — "the questions give the answer away", "two outcomes describe the same person" — and the AI writes another version from the same brief plus your correction. Be specific about what's wrong rather than what you want; "two outcomes overlap" produces a better second attempt than "make it better".

The new version arrives as a separate draft. The one you have stays exactly as it is, so you can read both and keep the better one. That is deliberate: a quiz with responses can't be rewritten in place without the answers already collected losing their meaning, and even on an untouched draft the second attempt might be worse.

Duplicate, optionally translated. From the quiz list, the ⋯ menu offers Duplicate or translate. A plain copy is a draft with its own URL and everything else identical. Pick a language and the copy is translated as well — only the words change, so the questions, outcomes, score bands, region bounds and every scoring weight come across untouched and nothing has to be re-balanced. Everything you wrote comes across: questions and options, outcome names and result copy, your email-gate wording, your challenge invite and the label on your result button.

Two things to check afterwards, because a translation can be correct and still land wrong. Outcome names carry the most weight for the least words — they're what a respondent says out loud and what gets shared — and a word with two senses can come out in the wrong one. And a result CTA's link keeps whatever tracking parameters the original had, so if you tag campaigns by language, update them on the copy.

This is also how you change a quiz's language: it's locked after generation, so you copy it into the language you want and keep the work instead of starting over.

One thing translation can't do: if a quiz's answers differ from each other by a hair — a spelling or grammar trap, where the wording is the question — the options collapse into the same sentence in another language and the answer key stops meaning anything. Snacked warns you when a quiz looks like that. Write a fresh one for the other language instead.

Theme and brand your quiz

Every quiz has its own theme so a single account can host multiple brands. The Theme tab gives you:

• Four built-in presets: Sprout (mint + warmth), Midnight (deep ink, high contrast), Sunset (coral gradients), Mono (black on white). One click and the entire quiz repaints.
• Custom colours. Set your primary, background and accent colours. The result page picks up the same palette.
• Custom font. Pick from a curated list of Google Fonts, or paste your own font URL.
• Logo. Upload your logo for the quiz header.
• Social preview. Upload a custom Open Graph image for when the quiz URL is shared on social.

Free accounts always show a small "Made with Snacked" badge on the result page. Pro removes it.

Publish and share

When you're ready, hit Publish. The quiz becomes live at a public URL:

snacked.it/<your-handle>/<your-quiz-slug>

Before you can publish, Snacked checks a short list: at least one question, at least two outcomes (score bands or regions, depending on the format), at least one scoring weight, and a verified email address. The list sits next to the Publish button, so you can see what's missing before you press it.

Your handle is assigned when you sign up — it's the snacked.it/<your-handle> part, and it's the address of every quiz you publish plus your profile page. Choosing your own is part of Pro, subject to it being free like any username, and a short list of words is reserved (route names like pricing or blog, and generic ones like info).

A handle you've chosen stays yours if you go back to Free. It's the address of pages other people have already linked to — newsletters, embeds on their own sites, messages already sent — so taking it back would break them, not you. What returns to Pro-only is the ability to change it again.

If you do change it on Pro, know what it costs: nothing redirects the old handle. Every link already shared stops working — public pages, the badge on results people have already seen, and any embed sitting on someone else's site. Snacked asks you to confirm and tells you how many quizzes will move.

The per-quiz slug is editable from Settings → URL. Edits to a live quiz go public as soon as they save — there is no separate "push changes" step, and the header says so while a quiz is live.

Once it's live, the Published menu gives you two ways to take it back:

• Pause — visitors who follow the link see a short "this one is closed" page instead of the quiz. Responses already collected are untouched, and you can put it back online whenever you like.
• Unpublish — back to draft. The public URL stops working entirely, so a link already shared leads nowhere.

Share the URL in your bio, in newsletters, in social posts. Or embed it — see Embed your quiz anywhere below.

Turn a result into an invite

A result is a full stop unless there's a reason to pass it on. Challenges give one: whoever finishes your quiz can create their own link, and everyone who answers from that link lands on the same board — a head-to-head at two people, a leaderboard at five.

The board sits at snacked.it/<your-handle>/<your-quiz-slug>/g/<id> and takes the shape of whatever your quiz measures: a ranking for a scored test, one map with everybody's dot for a compass, groups by outcome for a personality quiz, and the questions the group splits on hardest for a survey. Whoever is looking also sees how many answers each other person gave the same way they did.

Two things it deliberately does not do. The invite never reveals what the sender got — that is the thing the other person answers to find out. And nobody gets notified: a respondent stays anonymous, so the board simply updates at the link that was already shared, and whoever arrives second is the one who sends it back into the conversation.

The only thing Snacked asks for is a first name, and only from someone who creates or joins a challenge — after the result, never before it, and never an email.

Settings → Sharing holds four controls:

• Let people challenge someone — on by default.
• Who they invite — your word for the other person: a friend, your partner, a colleague, your team, your class. It goes on the button, and it is what makes the same mechanic fit a couples quiz and a team onboarding.
• The invite message — what travels with the link.
• Only the people in it can see a board — off by default, because a board anyone can open is a board that can be forwarded, which is how it grows.

The Analytics tab counts the whole loop: challenges started, invites opened, how many of those turned into answers, board views, and people per board. Challenges are on every plan, free included.

Collect emails

A quiz can ask for an email before it shows the result. A response that gives one is a lead: it lands in the quiz's Responses tab with the outcome, the answers and the UTM parameters attached.

The gate is off on every new quiz — you turn it on in the editor under Settings → Email gate, where you choose:

• Email collection — Required (no email, no result), Optional (a skip link) or Don't ask.
• Position — before the quiz, after N questions, before the result (the default and the one that converts best), or after a teaser of the result.
• The copy — headline, subtitle and button. This is the highest-leverage copy in the whole funnel: say what they get, not that you need their email.

Asking earlier gets fewer emails from more motivated people; asking at the result gets more emails from people who already invested the effort. If you have no strong reason, leave it before the result.

What it costs you. Every gate turns some respondents away, and a quiz you're running to be shared is usually better off with no gate at all. Gate the quizzes that feed your business; leave the ones that feed your reach open.

On the free plan you keep the first 25 emails, counted across all your quizzes, ever. After that the gate keeps working and the leads keep arriving — you'll see them in the list with the address masked and a Locked badge. Nothing is discarded: the moment you upgrade, every locked lead is unlocked at once, the backlog included, and you get an email confirming how many came back.

Connect your email tool

Integrations are a Pro feature, all of them. Open Settings → Integrations to connect:

• Klaviyo — list + per-outcome tags
• ConvertKit — sequence + per-outcome tags
• Mailchimp — audience + per-outcome tags
• MailerLite — group + per-outcome tags
• Brevo — list + per-outcome tags
• Zapier — every captured lead fires a webhook into your Zap
• Custom webhook — every captured lead POSTs a JSON payload to a URL you control

Why all of them are Pro, webhooks included: an integration ships a lead somewhere else. Free accounts do collect leads — they just read and export them by hand, in the dashboard. What Pro buys here is the automatic hand-off (and a locked lead is never dispatched, on any plan). Each captured lead sends the email, the outcome, the full set of answers, the quiz ID, and any UTM parameters present at capture. The outcome is sent both as a string and as a tag, so segmenting your list is one filter.

Add branching logic

Pro feature. From the editor's Branching tab you can attach rules to any option:

• Jump to question (skip ahead)
• End the quiz (go straight to a specific outcome, e.g. a disqualifier path)

Branching is useful for two cases. First, qualifiers: ask "Do you have a website?" and route "No" straight to a starter-tier result. Second, personalisation: skip irrelevant questions for users who already self-identified earlier.

Keep branching minimal at first. A 6-question linear quiz with a tight scoring matrix beats a 14-question branching maze.

Embed your quiz anywhere

Available on every plan. The Embed page on any quiz gives you three snippets:

• Inline iframe — paste into any HTML, sized to your container
• Popup — a button anywhere on your site opens the quiz in a modal
• Floating button — a corner button persistently invites visitors to take the quiz

All three pass UTM parameters from the host page through to the response record, so attribution survives the embed. On the free plan the embedded quiz carries the "Made with Snacked" badge; Pro removes it.

Track conversions

In Settings → Tracking on any quiz, paste:

• Meta Pixel ID — page views, quiz starts and lead events fire client-side
• Conversions API token — the same lead events fire server-side with event deduplication, so iOS 14+ and ad-blocked users still count
• GA4 measurement ID — Google Analytics 4 events for the whole flow

Pro feature. Add the IDs once, see attributed leads end-to-end. No tag manager required.

Read and export responses

Every completion creates a response, listed under the quiz's Responses tab. For each one you see:

• The outcome it resolved to
• All answers
• Completion time
• UTM parameters at capture
• The email, if the response was identified — masked while that lead is locked

Responses without an email show as Anonymous: that's every response on a quiz with no gate, and the ones who skipped an optional gate. The Analytics tab aggregates them: views, starts, completion rate, drop-off per question, outcome distribution and answer distribution per question — plus the challenge funnel once somebody starts one, see Turn a result into an invite.

Traffic (Pro). Analytics also breaks visitors down by where they came from and what they were using: views, starts, completions and leads for each source and each device class. A visit is attributed to its utm_source, or to the referring site when the link carried no UTM, or to direct. The column that earns the tab is views — it counts everyone who opened the quiz, including the people who read the intro and left, so you can tell a source that sends the wrong people from a quiz that loses the right ones. Traffic recorded before this existed has no landing data and doesn't appear.

On the free plan the page also shows how many of your 25 free emails you've used, and how many are waiting locked.

Export the full list as CSV from the Export button. The CSV includes anonymous responses; the email column is filled only for identified ones, and reads masked for locked leads. Free exports are capped at 100 rows per file.

Plans, limits and billing

Billing is handled through Paddle as Merchant of Record. Upgrade and downgrade any time from Settings → Billing. Cancellations take effect at the end of your current billing cycle — you keep Pro features until then. Customers in the EU/EEA/UK/Switzerland are entitled to a full refund within 14 days from the date of purchase per Paddle's Refund Policy.

Languages

The AI follows whatever language you pick at generation time, and everything around it follows too: the start button, the score and result labels, the email gate, the survey thank-you, the error messages, the challenge invite. Not just the questions and outcomes you generated — the whole page a respondent reads.

Italian, English, Spanish, French and German are translated in full. A quiz in any other language keeps its own generated copy in that language, and the labels around it fall back to English.

The dashboard UI is English for now. You can have multiple quizzes in different languages on the same account.

Troubleshooting

Generation failed midway. The scoring matrix integrity check rejected the output. Just hit Regenerate — failures are rare and idempotent.

The AI ignored part of my brief. Try one of two things: tighten the brief (shorter, more specific) or generate, then edit. The editor is the source of truth — the AI is the rough draft.

My ESP isn't connecting. Double-check the API key permissions. For Klaviyo and Brevo, the key needs at minimum the list/contact write scopes. If it still fails, send the error message to support@snacked.it — we keep a per-provider runbook.

The embed iframe is too tall / too short. The quiz auto-resizes via postMessage to the parent. Make sure your container doesn't have a fixed height; let it grow.

Leads are coming through but not tagged by outcome. Re-check that the outcome field is mapped in your ESP integration setup. Some providers (notably Mailchimp) require the destination tag field to exist before sync starts.

I have responses but no leads. Leads only exist when an email gate is on, and the gate is off by default on every quiz — turn it on in Settings → Email gate. If leads are arriving but the addresses read masked, you've passed the free plan's 25: they're locked, not lost, and upgrading reveals all of them.

What's next

• Take the live demo quiz — it tells you which of the four formats to build, and it was built with Snacked
• See pricing — Free and Pro, side by side
• Read the FAQ for shorter answers to common questions
• Email support@snacked.it for anything we haven't covered

Snacked is built by a small team that ships fast. If something feels off, tell us — most weeks at least one user-reported friction makes it into the next release.