Riddle Technologies today published The State of Interactive Content 2026, a white paper built on more than 3.8 billion views of quizzes, polls, personality tests, forms and mini-games embedded on publisher and brand sites between June 2025 and May 2026. The Saarbrücken company leads with an average of 112 seconds of active engagement per start. The finding with wider consequences for measurement teams sits near the end of the document: nearly four in five shares of those units travel through private messaging apps that referral reports cannot attribute.

In Short

A company that makes quizzes and polls for websites looked at a year of its own data and found that people who start one spend almost two minutes on it, and most of them finish. That matters to news sites and brands because fewer visitors now arrive from Google, so holding people on the page and collecting their email addresses has become more valuable. The catch is that most sharing happens in WhatsApp and other private apps, so the traffic numbers you see undercount how far this content really travels.

What the dataset covers

According to Riddle, the report draws on aggregated, anonymised interaction statistics from its customer community. Every figure is expressed as an aggregate or a rate, and no individual user records are included. The window is the trailing twelve months from June 2025 to May 2026, which the company says it chose so that the 2026 label reflects current behaviour. Bot traffic and sessions shorter than three seconds are stripped out of the engagement figures.

Four definitions govern everything that follows. A view is a unit loaded or an impression served, and it is the basis for the 3.8 billion headline. A start is recorded when a user answers or advances past the first screen. A completion means the user reached the final screen. Time per start is the average active engagement time, in seconds, among people who started. Throughout the document, "a Riddle" is the company's catch-all term for any unit built on its platform, which it says offers more than 35 formats; figures are broken out for only five.

The company calls the sample representative of activity across a community spanning more than 20 industries. It does not say how the sample was drawn, how many accounts or sites contributed, or how traffic splits by country. That matters because the document is also a sales asset. Its final page carries a demo booking prompt and a QR code, which places it in the long tradition of the vendor white paper, where the publisher of the data is also the seller of the product the data supports.

Attention per start, and the Chartbeat comparison

Over the twelve months, Riddle counts more than one billion active starts and an overall completion rate of 72.7%. Average time per start came to 112 seconds, which the company totals at 31.3 million hours of engagement, or more than 3,500 years. The arithmetic is consistent. One billion starts at 112 seconds each gives roughly 31.1 million hours, and the report's figure rests on a start count described only as "1 billion+". Divided by the hours in a year, 31.3 million hours comes to about 3,570 years.

The report calls the 112 seconds "the single most important number for a marketing audience" and sets it against Chartbeat's Average Engaged Time, which Riddle cites at 26 seconds globally for the second quarter of 2026. According to Riddle, Chartbeat measures engaged time actively, through scrolling, typing or mouse movement in a foreground tab, on the same basis as its own metric. The white paper describes the gap as three to four times. On its own numbers the ratio is closer to 4.3, so the stated range is, if anything, conservative.

The denominators differ, however. Chartbeat's average is taken across pageviews on its publisher network. Riddle's is taken only among people who have already chosen to interact, after sessions under three seconds have been removed. The report's own format table puts the volume-weighted view-to-start rate at 43.0%, which means more than half of all loaded units were never started and contribute nothing to the 112-second figure. The benchmark itself also moves by market: PPC Land's coverage of the same Chartbeat quarter recorded engaged time peaking in Central and Eastern Europe at 38.4 seconds. A comparison between an opted-in audience and an all-comers average flatters the former by design.

Five formats, five different jobs

The format breakdown is the most granular section of the report. Completion appears as a median rather than a mean, which Riddle says is more representative for engagement data, and volume is expressed as each format's share of total starts.

FormatView to startMedian completionAvg time per startShare of starts
Quiz48.6%72.9%135 s~73%
Poll18.1%98.6%12 s~19%
Personality test52.7%84.6%97 s~7%
Mini-game33.4%33.7%396 s~1%
Form23.4%80.0%19 s<1%
Weighted average43.0%78.0%112 s100%

Quizzes carry the volume. Nearly half of the people who load one begin it, and those who begin stay for more than two minutes. Personality tests hold the highest view-to-start rate at 52.7%, which the report attributes to the curiosity of a "which one are you?" prompt. Polls post a 98.6% median completion, though Riddle is candid that the figure is structural: polls are almost always single-question units, so casting a vote is itself a completion. Twelve seconds per start confirms how thin that engagement is.

Mini-games are the outlier on time. At 396 seconds, about 6.6 minutes, they hold attention roughly 2.9 times as long as a quiz and 33 times as long as a poll. Their 33.7% completion rate looks weak until the report's explanation is taken into account: open-ended play, such as Sudoku or crosswords, does not always have a defined finish. Mini-games are also Riddle's newest format and account for about 1% of starts, a share the company expects to rise as adoption spreads. That is a forecast from the vendor, not a measured trend.

One reconciliation is missing. The weighted median completion in the format table is 78.0%, while the headline overall completion rate is 72.7%. The report does not explain the difference. The most likely reason is methodological, since one figure is a volume-weighted blend of per-format medians and the other appears to be a pooled rate, but the document does not say so.

Mike Hawkins, co-founder at Riddle, framed the findings around retention rather than reach. "The challenge for publishers is no longer simply getting people to a page. It is creating an experience compelling enough for them to stay, participate and return," he said. "Our data shows that interactive content can turn attention into an active relationship - while also creating opportunities for first-party data and organic distribution." The report contains no data on return visits, so the third verb in that sentence remains a claim rather than a finding.

The first question carries most of the loss

Aggregated across formats, the average submit rate by question position runs as follows: 81.9% at the first question, 94.2% at the second, 96.3% at the third, 97.8% at the fifth and 98.2% at the tenth. About 18% of starters leave at the opening question. After that, attrition flattens sharply. Multiplying the first three rates together gives a cumulative pass-through of roughly 74%, close to the headline completion rate.

Rising submit rates deserve a caution. Only people who stayed reach later positions, and a tenth question exists only in longer units, so the curve partly reflects who is left rather than how compelling later questions are. Riddle reads the pattern as momentum. Selection explains at least some of it too.

Length shows a similar but less tidy picture. Units with one to five questions posted a median completion of 86.4%. Six to ten questions fell to 75.2%, eleven to fifteen rose slightly to 77.5%, and sixteen or more dropped to 59.1%. The report does not address why the eleven-to-fifteen bracket outperforms the shorter six-to-ten bracket. Riddle's own takeaway is that experiences aimed at completion stay at five questions or fewer, with longer formats reserved for high-intent audiences.

Vertical results, and a table that does not add up

Riddle consolidated its more than 20 industries into ten verticals and ordered them, by its own description, from longest to shortest engagement time, with the catch-all Other group placed last.

VerticalAvg time per startCompletion rate
Education & Learning212 s71.3%
Professional & Financial Services158 s72.3%
News & Publishing112 s73.4%
Entertainment & Gaming103 s76.5%
Non-Profit & Public Sector94 s75.9%
Agencies92 s71.4%
Health & Wellness81 s78.9%
Sports70 s74.6%
Retail & Travel68 s69.3%
Other212 s64.6%

Three inconsistencies appear in this section. The Other row shows 212 seconds, identical to Education & Learning, which would make it joint longest in a table sorted by time; the figure may be a transcription error, but the document does not clarify. The accompanying commentary describes the education sector's engagement as 150 to 210-plus seconds, a range that does not match the single 212-second value in the table. And the report says News & Publishing "dominates engagement volume", yet it publishes no volume breakdown by vertical to support that statement.

On the figures that are consistent, Health & Wellness leads on completion at 78.9%, with Entertainment & Gaming at 76.5% and Sports at 74.6%. Retail & Travel sits at the bottom of the named verticals on both measures, at 68 seconds and 69.3%. News & Publishing lands exactly on the platform average of 112 seconds.

The vertical split overlaps with where search losses have concentrated. Index Exchange data covering 1,200 publishers showed health and education among the hardest-hit categories for ad opportunities while news held up comparatively well. In Riddle's data, education produces the longest sessions and health the highest completion. The two datasets measure different things, and no causal link is drawn in either, but the overlap describes the verticals with the most to gain from holding the visitors they still receive.

Where the lead form sits

The lead generation section is the most operationally specific part of the report. Across lead-enabled units in the sample, Riddle compared three placements of the opt-in step. Opt-in rate is defined as the share of starters who went on to submit the form, taken as a median across units. The company says a start-based denominator keeps the placements comparable.

Lead-step placementView to startStart to completionOpt-in (share of starts)
Start, before first question4.5%78.0%92.2%
Middle16.3%73.9%54.6%
End, after last question43.7%51.9%26.9%
Overall14.0%61.0%61.0%

Front-gating turns the form into a barrier. Only 4.5% of viewers pass it, but 92.2% of those who do opt in. With end-placement, the form is invisible to someone deciding whether to begin, so 43.7% of viewers start. By then only 26.9% of starters submit their details. Riddle notes that the 43.7% is an aggregate across all formats in the end-placement cohort, sitting above a simple five-format average of 35.2% because that average underweights quizzes.

One figure goes unexplained. Units with the form at the end show a start-to-completion rate of 51.9%, well below the 72.7% platform average. The report does not say whether the form screen sits before the final result, which would make the form itself an exit point, or whether the cohort differs in some other way.

The 10,000-view model

Riddle applies each placement's rates to a hypothetical 10,000 people who load a unit. Front-gating yields 450 starts, 351 completions and 415 leads. Middle placement yields 1,630 starts, 1,205 completions and 890 leads. End placement yields 4,370 starts, 2,268 completions and 1,176 leads. The overall row gives 1,400 starts, 854 completions and 854 leads.

End placement therefore captures nearly three times as many leads in absolute terms, 1,176 against 415, because almost ten times as many people begin. Front-gating is the only placement that produces more leads than completions, since the email is collected before the content and some people leave after handing it over. Riddle frames the choice as a trade between intent and volume.

The model is an illustration, not a count of observed leads. It multiplies median rates that were calculated separately across different units, and medians of separate ratios do not necessarily combine into a coherent funnel. The overall row, where completion and opt-in are both exactly 61.0%, is left without comment.

The broader data-collection context is familiar. Survey work on content marketing from the Content Marketing Institute found gated assets, webinars and interactive tools accounting for 68% of first-party data collection among the marketers it surveyed. Riddle positions its platform as compliant with the GDPR and, per its company description, ISO 27001 certified. The white paper does not present evidence for either claim, and a form submission counted as a lead is not the same thing as consent for every later use of that address.

Sharing that referral reports cannot see

Riddle measures share rate as shares per completion. Personality tests lead at 0.66%, more than double quizzes at 0.30%, followed by mini-games at 0.22%, polls at 0.20% and forms at 0.14%. In absolute terms these rates are small: even the most shared format generates roughly one share per 150 completions. The report attributes the personality test lead to the social nature of a result that people want to show others.

The channel data carries the measurement point. WhatsApp accounts for 75.0% of tracked shares, Facebook for 12.4% and X for 9.2%. Email follows at 1.1%, Messenger at 0.9%, LinkedIn at 0.7%, copied links at 0.4%, SMS at 0.2% and Telegram at 0.1%, with Instagram, Viber, Pinterest, Reddit and others together under 0.1%. Riddle classifies WhatsApp, Messenger, email, SMS, Telegram and copied links as dark social. Summed, those private channels come to 77.7% of shares, the basis of the company's rounded headline figure. Facebook, X and LinkedIn, the open networks, make up 22.3%.

According to Riddle, visits generated by private channels usually arrive with no referrer and are logged as direct traffic. The report concludes that on-platform share counts "materially understate true reach". That conclusion runs in one direction only, since the dataset counts shares initiated through the unit's own share controls. Links copied from a browser bar or passed on as screenshots never register, so the private share of distribution may be higher still.

The attribution gap is not new, and it is widening from several directions. Similarweb panel research found only 8.8% of AI-influenced visits arriving through a direct AI referral, with most routed through branded search. Google moved in May 2025 to fix an AI Mode bug that stripped referrer data and caused clicks to register as direct, and Google Analytics added a dedicated AI Assistant channel in May 2026 after years in which such sessions landed in Referral or Direct. No equivalent default channel exists for a link forwarded in a WhatsApp group.

Why the report lands now

Riddle's framing leans on a traffic shift PPC Land has tracked throughout the year. Chartbeat data reported in March 2026 showed small publishers losing 60% of search referral traffic over two years, against 47% for medium publishers and 22% for large ones, with Google Search pageviews down 34% between December 2024 and December 2025. A randomised field experiment involving 1,065 desktop Chrome users found that AI Overviews cut outbound organic clicks by 39.8%and raised zero-click searches by 34.5%.

Publishers are paying to replace some of that traffic. The hundred largest media publishers spent an estimated $113 million on paid search in July 2026, 41% more than a year earlier and 274% more than three years before, according to Similarweb tracking reported by Adweek. Others are rebuilding around the session rather than the pageview. Teads built its EngageOS product around revenue per session in June 2026, and JWX claimed in August that search referral had fallen from as much as 70% of publisher visits to under 20%, a figure published without a dataset.

Riddle's white paper belongs to that same group of vendor arguments. Each company selling on-page engagement tools has an interest in showing that engagement can substitute for lost clicks, and each publishes its own figures to make the case. Riddle's contribution is more detailed than most, with format, funnel, vertical and placement data drawn from a large base. What it does not contain is the connection that would matter most to a publisher's finance team: whether 112 seconds inside a quiz translates into ad yield, subscriptions, return visits or revenue. The report offers no outcome data of that kind, no year-over-year comparison despite referencing trend context back to January 2024, and no third-party audit of its figures.

For advertisers, the implications are narrower but concrete. Interactive units sit on publisher pages that also carry display and video inventory, and a visitor held for nearly two minutes is a visitor exposed to more of it. Whether that exposure is measured, and whether the reach generated through private sharing ever reaches a planning tool, depends on measurement infrastructure the report itself describes as blind to most of that distribution.

Timeline

Summary

Who: Riddle Technologies, a Saarbrücken-based no-code platform for quizzes, polls, personality tests, forms and mini-games embedded on publisher and brand websites. Co-founder Mike Hawkins is quoted. The findings concern publishers, media companies, brands and the advertisers buying inventory on their pages.

What: The State of Interactive Content 2026, a white paper reporting 3.8 billion views, more than one billion starts, 112 seconds of average engagement per start, a 72.7% overall completion rate, format and vertical breakdowns, lead-form placement trade-offs ranging from a 92.2% opt-in rate when gated at the start to 43.7% start rates when the form sits at the end, and sharing data showing WhatsApp carrying 75% of tracked shares and private channels 77.7% in total. The vertical table contains unreconciled figures, including an Other row matching Education's 212 seconds.

When: Published today, with a measurement window running from June 2025 to May 2026.

Where: Across Riddle's global customer community in more than 20 industries, consolidated into ten verticals. The report does not disclose a geographic breakdown.

Why: Search referrals to publishers have fallen sharply as AI-generated answers absorb clicks, and vendors selling on-page engagement tools are presenting data to show that holding visitors and collecting first-party data can offset part of that loss. The report's sharing data also indicates that standard referral reporting misses most of the distribution these units generate, while offering no evidence linking engagement to revenue.