96% of U.S. adults say they use the internet, according to the Pew Research Center survey of 5,022 adults fielded between February and June 2025. That number makes the online questionnaire look like a way to reach everyone. The same fact sheet also reports that 16% of U.S. adults are smartphone-only internet users, and that is where a survey design starts to matter.
An online survey works as an empirical research method when your research question asks about attitudes, opinions or self-reported behavior, when your target population is reliably reachable online, and when you want to test a relationship between variables rather than estimate a value for an entire population. By the end of this article you will have a three-sentence template for the limitations section of your thesis.
📌 Key takeaways
- 96% of U.S. adults used the internet in 2025.
- Among adults aged 65 and older the figure is 90%.
- 16% of U.S. adults reach the internet only through a smartphone.
- AAPOR requires you to state which sample type you used.
- Margin of error does not apply to non-probability samples.
Create a survey for free
With empirio.ai you can create a modern online survey in minutes — free to start.
- AI-built survey
- Adjust by drag & drop
- Real-time analysis
What makes an online survey a research method
An online survey is a standardized form of questioning in which the respondent completes the questionnaire on screen, with nobody reading out the questions or recording the answers. Methods texts call this a self-administered survey, and that single feature explains both its strengths and its weaknesses.
The absence of an interviewer buys you consistency. Every respondent sees the same wording, the same order and the same answer options, and no tone of voice nudges a rating in one direction. Nobody can rephrase a question on the spot for one person and not for another, which is exactly what standardization is supposed to guarantee.
What you give up is control of the setting. You do not know which device someone is using, how distracted they are, or whether the person you meant to reach is the person answering. Misunderstandings stay invisible as well: a question an interviewer would clarify in ten seconds becomes an answer that looks exactly like every other answer in your dataset.
One thing an online survey does not decide for you is the direction of your research. An online questionnaire is a mode of data collection, not a research tradition. Whether your study is qualitative or quantitative depends on your questions and your planned analysis, not on the tool.
Empirical research methods at a glance
Methods texts usually distinguish five empirical research methods: surveys, observation, experiments, content analysis and secondary analysis. The online questionnaire belongs to the first of them.
| Method | What it gives you | Typical form |
|---|---|---|
| Survey | self-reported opinions and behavior | questionnaire, interview, focus group |
| Observation | actual behavior rather than self-report | participant or non-participant, overt or covert |
| Experiment | evidence about cause and effect | laboratory experiment, field experiment |
| Content analysis | patterns in texts, images and posts | coding against a category system |
| Secondary analysis | answers from somebody else's dataset | reanalysis of existing studies |
Two distinctions are worth getting right in a defense or a committee meeting. An experiment is strictly a research design rather than a data collection technique, because the data in an experiment still arrive through a survey, an observation or a measurement. Secondary analysis collects nothing at all; it reworks datasets that other researchers gathered.
Within survey research, the dividing line is the degree of standardization. A structured questionnaire gives you countable answers from many people, a semi-structured interview gives you few answers in depth, and a focus group lives on what happens between the participants. Experiments add the question of setting: a purpose-built environment makes it a laboratory experiment, the participants' own everyday life makes it a field experiment. For how these fit together in a project, see our articles on data collection and on empirical research.
When an online survey is the right choice
Whether the online mode fits is decided by your research aim, not by your topic. The USC Libraries guide to research designs states the rule behind that in one line: the research problem determines the type of design you choose, not the other way around.
In practice the dividing line runs between two jobs: testing a relationship and estimating a value for a defined population. In a comparison, random assignment does the work. If you want to know whether version A of an ad is rated more highly than version B, who happens to be in your sample matters far less, as long as the two versions were assigned at random. If you want to say what percentage of U.S. college students do something, you need a selection that mirrors that population, and a shared link will not give you one.
| Your research aim | Online survey | Why |
|---|---|---|
| Test version A against version B | well suited | random assignment carries the comparison |
| Test a relationship between variables | usually suited | unless self-selection is tied to the topic itself |
| Estimate proportions for a population | only with a probability sample | no list of all internet users exists |
| Survey every member of one organization | well suited | if you can access the full mailing list |
| Reach a group with low internet use | not suited | too many people are excluded by the mode |
For most graduate projects that is good news. A master's thesis is not trying to measure the United States; it is trying to show a relationship, and an online questionnaire does that job cleanly. All your text has to do is say which of the two aims you set yourself.
Coverage: who you cannot reach online in the United States
A coverage problem arises when part of your population, meaning everyone you want to make a statement about, cannot take part for technical reasons. Someone who is not online cannot answer, however well your questionnaire is written.
The Pew Research Center internet and broadband fact sheet puts adult internet use at 96% in 2025 and shows where the remainder sits. Adults aged 18 to 49 are at 99%, those aged 50 to 64 at 96%, and those aged 65 and older at 90%. Alongside that, 78% of U.S. adults subscribe to home broadband and 16% are smartphone-only users.
That last figure matters more for a student project than the coverage gap itself. A questionnaire with a wide grid, a long matrix question or a file upload is far harder to complete on a phone, and the people who drop out are disproportionately the ones with no other device. Coverage and usability push in the same direction here.
⚠️ Warning
The narrower your target group, the less a national average tells you. Check instead whether your specific group is reachable through the channel you planned, for example shift-working nurses or adults over 75. If it is not, run a paper version alongside the online questionnaire and describe both routes in your methods section.
What representativeness in an online survey actually requires is set out in a separate article.
Create a survey for free
With empirio.ai you can create a modern online survey in minutes — free to start.
- AI-built survey
- Adjust by drag & drop
- Real-time analysis
Opt-in samples and what AAPOR requires you to report
An opt-in sample is one in which respondents decide for themselves whether to take part, because you shared an open link. In student research that is the normal case, and it is not a flaw as long as you report it as one.
The AAPOR Code of Professional Ethics and Practices, revised in June 2026, does not prohibit opt-in samples. It requires disclosure. Researchers must explicitly state whether the sample came from a probability-based frame, meaning potential participants selected with a known non-zero probability from a known frame, or from non-probability methods such as opt-in or volunteer sources.
The second AAPOR rule is the one that catches most student papers. Estimates of sampling error, usually described as the margin of error, are reported for probability samples. For non-probability samples, AAPOR asks researchers either to describe in detail the model behind any measure of precision, including how its assumptions were validated, or to state plainly that the sample was selected with a non-probability method and that measures of precision are not provided. A margin of error printed under an opt-in online survey is not a small imprecision; it is a claim the design cannot support.
AAPOR binds its members, not students, and your program requirements remain the authority for your own work. For the recruitment routes themselves, see our article on finding survey participants; the selection procedures and the question of size are covered in the article on sampling.
Advantages and disadvantages of an online survey
The advantages and disadvantages of an online survey fall into two groups: the practical ones, which decide your timeline, and the methodological ones, which decide how much weight your results can carry. Your methods section is about the second group.
What the method gives you
- Every respondent sees exactly the same questionnaire.
- No interviewer tone or follow-up shifts an answer.
- Sensitive questions attract fewer socially desirable answers.
- Skip logic and rotation stay invisible to the respondent.
- Answers reach the dataset without transcription errors.
What you accept in return
- Respondents decide for themselves whether to take part.
- Misunderstood questions are never flagged to anyone.
- Faults in the skip logic go unnoticed during fieldwork.
- Repeat entries are hard to rule out with an open link.
- Without a mailing list there is no real response rate.
Cost, fieldwork time and the effort of building the questionnaire are compared in detail in our article on the online survey and its pros and cons. Where a mailing list does exist, the figure is a real one, and you can calculate the response rate from invitations sent and responses received.
How to justify your method in the methods section
A methods section answers four questions in this order: what did you want to know, how did you collect it, whom did you ask, and where do your claims stop. The fourth is the one most often left out, and the one a committee asks about most often.
- State the research aim. Say in one sentence whether you are testing a relationship or estimating a value. That distinction carries the whole justification.
- Name the method and rule out alternatives. Why a survey rather than observation, why online rather than paper or telephone. Half a sentence per rejected option is enough.
- Describe recruitment and sample. Which channels, which period, how many views, how many usable cases, how many drop-outs. Numbers belong here, not in the conclusion.
- Set out the limitations. A paragraph of its own saying who is missing and what your results therefore cannot show.
💡 Tip
For the limitations paragraph, this template will get you most of the way: “Participation was through an openly shared link on [channel], so the sample is self-selected. Random selection from the target population was not possible, as no complete list of the target individuals exists. The findings therefore describe the sample surveyed and cannot be generalized to [population].” Three sentences, and the most common criticism is dealt with.
One step usually comes before all of this. If your study collects information through interaction with living individuals, it involves human subjects as defined in 45 CFR 46.102(e)(1), and your institution decides whether it needs review by an Institutional Review Board. Ask your IRB office before fieldwork rather than after. Where the methods section sits in the overall structure is covered in our article on the empirical research process.
Common mistakes with online surveys as a method
The three most common mistakes with an online survey as a research method happen at the writing stage, not during fieldwork. The questionnaire was fine, the data are in, and the text still claims something the study cannot support.
The first two mistakes share a root: the text promises more than the data can carry. The third is the reverse, where the text says less than the researcher actually thought through.
Mistake 1: calling the sample representative
A large number of responses reduces random variation but corrects no bias. A sample becomes representative through random selection from a defined population, not through size. Matching the age and gender profile creates agreement on those two characteristics and says nothing about any other. Write “the sample surveyed” rather than “students”.
Mistake 2: reporting a margin of error you cannot compute
A margin of error assumes a probability sample. With an opt-in link there is no known selection probability, so the figure a statistics package prints is not a margin of error for the population, only a description of the people who answered. AAPOR asks you either to document the model behind any precision estimate or to say that measures of precision are not provided.
Mistake 3: describing the method instead of justifying it
“An online survey was conducted” is a description, not a justification. The justification comes from the options you turned down: a semi-structured interview would have given deeper answers but not enough cases for the planned group comparison. Two sentences like that show the choice was deliberate, and that is exactly what a committee looks for.
Conclusion
An online survey is only as good as the question you ask of it. For comparisons and tests of relationships its sampling problems weigh less heavily; for statements about an entire population it needs a probability sample from a defined frame, and that is almost never available in a thesis or dissertation. Naming that limit yourself does not weaken your work; it makes it checkable.
Where to go next
- Ready to build the questionnaire itself? How to create a questionnaire
- Writing an empirical thesis right now? Empirical thesis: how it works
- Already planning the analysis? Methods of analysis at a glance
Method justified, questionnaire still to build?
With empirio.ai, an online survey tool from Germany, you build your questionnaire in a few steps, share it by link or QR code and watch responses come in live, so you can adjust while fieldwork is still running.