Online Survey: Definition, Pros, Cons and When to Use It

96 percent of U.S. adults say they use the internet, which leaves 4 percent who do not (Pew Research Center, 2025, fielded February to June 2025 among 5,022 adults). Whether that gap ruins your survey or matters not at all depends entirely on whom you want to ask.

An online survey is a standardized survey in which participants complete a questionnaire themselves through a link in their browser, without anyone asking the questions aloud. Online surveys are quick, cheap and hand you the data in digital form right away, yet they only ever reach people who are online. By the end of this piece you can decide for your own project whether the online mode fits.

📌 The key points at a glance

• An online survey is a standardized survey run through a link.
• The upside: low cost, short field time, instantly digital data.
• The drawback: an online survey reaches only people with internet access.
• Four percent of U.S. adults do not use the internet.
• A large number of responses does not create representativeness.

What is an online survey?

An online survey is a standardized survey whose questionnaire is filled out over the internet: the person being surveyed opens a link in a browser, answers the questions with nobody else present and submits the form.

In the methodological literature the online survey goes by the abbreviation CAWI (= computer-assisted web interviewing). What that names is not the technology but the mode of data collection: whoever fills out the questionnaire does so without a counterpart and at a pace they set themselves. Both the strengths and the weaknesses of this form grow out of that single property, and neither can be had without the other.

Three terms get mixed up in everyday use, and a methods chapter keeps them apart:

An online survey is usually quantitative in design, because a standardized questionnaire puts the same questions to everyone in the same order. Qualitative access through open text boxes is possible, yet it replaces no interview. How the two approaches differ is set out in the piece on qualitative and quantitative surveys.

An online survey is not a separate type of questionnaire. Online is simply a way of delivering the same questionnaire.

What are the advantages of an online survey?

The advantages of an online survey lie in cost, speed and data quality: there are no interviewer fees and no postage, the field time shrinks from weeks to days, and the answers arrive as a finished dataset without anyone typing them up.

Wolfgang Bandilla names exactly these points in the GESIS Survey Guidelines (2015) as the reasons the online mode has prevailed: short field times, immediate storage of the answers, complex routing and the option of building in multimedia elements. Rather than a technical convenience, all of that is a methodological gain, because every working step that falls away is a step at which no error can arise.

What the online mode actually delivers

• The questionnaire is built once and completed any number of times.
• Filters and skip rules run automatically, so nobody lands in the wrong question.
• The answers are ready for analysis as a dataset right away.
• On sensitive topics people answer more openly with no counterpart present.
• Images, videos and sliders can be built into the form.

Losing the counterpart is the methodologically most interesting of these advantages. Because no interviewer sits in the room and no voice reacts on the telephone, the pressure to give a socially desirable answer falls away. On questions about income, health, conflict in a team or a person’s own behavior that effect is well documented, and for many projects it is the strongest reason to collect data online at all.

What are the drawbacks of an online survey?

The central drawback of an online survey is its access: only somebody who is online, who receives the link and who decides to take part can do so. Everything else, from drop-offs to unserious answers, follows from that one property.

What you take on in exchange

• Queries go unanswered, because nobody is there to explain a question.
• Who actually filled out the form cannot be verified.
• Drop-offs mid-questionnaire are more frequent than in supervised surveys.
• People without internet access or a suitable device fall out entirely.
• Without an invitation list, those addressed decide themselves whether to join.

Uncertainty about who really completed the form can be softened, never dissolved. Sending personalized links at least tells you which invitation was used, though not who sat in front of the screen. For a thesis or a quick pulse check that is generally enough, for a survey with legal consequences it is not.

⚠️ Watch out

A common mistake is the combination of an open link and the claim that the sample was drawn at random. Anyone who shares the link freely has self-selection and says so in the methods chapter. The way out is a fixed invitation list with personal links.

Online, mail, telephone or in person: which survey mode fits?

The four classic survey modes differ above all in whom they reach and what they cost: the online survey is the fastest and cheapest, the in-person interview the most expensive and most thorough, with the mail and telephone forms in between.

Bandilla (2015) recommends in the GESIS Survey Guidelines that where a target group has insufficient internet access, you switch to a mail or in-person mode, because neither carries a comparable barrier to access. Choosing a mode is therefore not a matter of taste but a consequence of whom you want to ask. The American Association for Public Opinion Research publishes the corresponding guidance for study design in its 2023 task force report on online samples.

For many projects the choice of mode is settled faster than it sounds:

Mixed mode means combining several modes inside the same study. The GESIS Panel, for instance, surveys its participants online and brings people who do not use the internet in by mail, so that both groups end up in the same dataset (Bandilla, 2015). For a thesis that effort is usually too great, for a government agency or a research institute it is the standard route.

When an online survey is the wrong choice

An online survey is the wrong choice as soon as a noticeable part of your target group is not online. Survey methodology calls that the coverage problem: the inadequate overlap between the intended population (= everyone you want to make a statement about) and the group that is able to take part at all.

Generally speaking, a national average helps little with that question. 96 percent of U.S. adults say they use the internet, which leaves 4 percent who do not (Pew Research Center, 2025). Spread across the population, though, that gap sits very unevenly, and the uneven spread is what decides your project.

Sources: Pew Research Center, Internet and Broadband Fact Sheet, published 20 November 2025, National Public Opinion Reference Survey fielded February to June 2025 among 5,022 adults, margin of error plus or minus 1.9 percentage points. The 2023 figures come from the National Telecommunications and Information Administration. As of August 2026, subject to change.

Read the two 2023 rows as a cross-check and not as the same measure. Pew asks adults whether they use the internet at all, while the National Telecommunications and Information Administration counts people aged 3 and over and separately counts households without a connection, which is why the numbers sit so far apart. Both describe a gap, and neither describes it in the same way.

A breakdown by age is where the evidence thins out, and the thinness is a finding in its own right. Pew does not break its 2025 figure down by age group in the fact sheet, so the most recent value that can be cited with a year attached comes from 2021: back then, 25 percent of adults aged 65 and over said they did not use the internet. Anyone surveying older or less connected groups therefore budgets for a second channel from the outset, instead of leaning on a single headline percentage.

The same arithmetic applies to any other subgroup. In a survey of people without a high school diploma, of people needing care or of residents in institutions, the share reachable online lies well below the national figure. Among students, office staff or club members under 60, the coverage problem plays practically no role.

Does a high number of responses make a survey representative?

No, a high number of responses does not make an online survey representative. A survey becomes representative through the way its participants are selected and not through its size: they have to be drawn at random from the population, otherwise even a large sample stays skewed.

Bandilla (2015) names in the GESIS Survey Guidelines the reason this is particularly hard online: for the population as a whole there is no complete, currently valid list of people with internet access, and therefore no list of email addresses from which a random sample could be drawn. Telephone surveys can generate random numbers artificially, whereas email addresses allow nothing comparable.

The American Association for Public Opinion Research arrives at the same place from the other direction. Its 2023 task force report on data quality metrics for online samples names self-selection and the coverage of people who do not use the internet as the quality problems of online panels. Nevertheless the online mode works well for experiments and hypothesis tests, because what counts there is the random allocation to experimental groups and not the selection of respondents.

💡 Tip

For a thesis a non-random sample is rarely a problem, as long as you name it. A sentence such as “The sample is a convenience sample, so the results cannot be generalized to the population” costs you no points, and it will not trip you up in the defense either. Concealing it costs you some.

How many participants you actually need, and what else representativeness hangs on, is set out at length in the piece on the representative survey.

Common mistakes in online surveys

Most mistakes in online surveys are made long before the analysis: in the way the target group is cut, the way the questionnaire is built and the way the invitation goes out. Four of them turn up especially often in theses and in professional practice.

All four can be fixed before the link goes out and not afterwards. So the half day that a proper pretest with five test readers costs is almost always worth it.

The questionnaire is unusable on a phone

A large share of responses now comes from smartphones, and rating grids with seven columns simply do not work there. The consequence is not an empty dataset but a poor one: somebody battling through an unreadable grid ends up clicking the middle column every time. Open every questionnaire once on your own phone before you send it.

The invitation promises anonymity while the tool still records

Anonymity is a setting in the survey tool, not a statement of intent in the invitation text. As long as email capture, a mandatory account or IP storage is switched on, the survey is not anonymous, whatever the invitation says. Which settings to check is covered in the piece on data privacy in surveys.

Anonymity is also not the only obligation, even though there is no GDPR in the United States. The closest equivalent is state law: under the California Consumer Privacy Act as amended by the California Privacy Rights Act, a business that hands personal information to a service provider needs a contract with that provider, which Cal. Civ. Code § 1798.100(d) requires. The California Attorney General keeps an overview of the CCPA, and for student work your advisor and the IRB set what applies to you.

The questionnaire is too long

Drop-offs are not spread at random: whoever quits differs systematically from those who see it through, and that difference distorts the result. Ten minutes is the upper limit for most projects, and closer to five where participation is voluntary and unrewarded. Cutting questions hurts less than a sample made up only of the unusually motivated.

The field time is set too short

A field time of 48 hours excludes everyone who is on vacation, on shift work or sick during those two days. What arises is the same systematic bias as with an overlong questionnaire, only less visible. Two weeks with one friendly reminder after the first week is a good rule of thumb and costs nothing but patience.

Conclusion

An online survey is the right choice for the great majority of projects, because it is fast, cheap and low in error, and it is the wrong one as soon as a relevant part of the target group is not online. Decide first whom you want to ask, and only then how. If you do collect data online, name the self-selection openly rather than stepping around it: a study that knows its limits is worth more than one that hides them.

Where to go next

• Ready to build the questionnaire? How to create a questionnaire: steps and structure
• Looking for a template to adapt? Templates and examples for surveys
• Wondering how many participants are enough? When an online survey is representative
• Want a better response rate? Tips for the perfect online survey
• Writing the methods chapter right now? Empirical research: definition, methods, guide

Want to launch your online survey today?

With empirio.ai, an online survey tool from Germany, you build the questionnaire in a few steps, keep the data hosted in the EU and analyze the answers inside the tool.

Start free with empirio.ai