Blog

Can you trust online reviews? How to tell who says they used it

You can trust an online review more when the person says they used the product and gives details an owner would know. Label each opinion by those signs, and keep the opinions without them as weaker evidence, not as noise.

Can you trust online reviews?

Some of them, for some questions. Shoppers have doubted online reviews for years. In 2016, Pew Research Center asked US adults about online reviews. It asked review readers whether reviews give a true picture of a product’s quality, and whether it is hard to tell if they are truthful.

A shopper asks, “Is this review real?” That question is hard to answer from the words alone. A research team can ask a narrower one: does this opinion come from someone who says they used the product? The words often answer that.

The answer changes what an opinion can support. A battery complaint from someone who wore a ring for three months is evidence about the battery. The same complaint from someone who read about it is evidence about what shoppers fear. Both are useful, but they answer different questions.

This guide uses one name for the stronger group: people who say they used it. The name states its own limit. A post can state use, but it cannot prove a purchase.

What are the signs that someone says they used the product?

Look at what the person says, not at how polished it reads. Strong opinions often carry several of these signs at once.

  • Words of own use. “I bought it in March.” “Mine stopped charging.” “I have worn it every night for a month.” The person states their use in their own words.
  • Time with the product. They say how long they used it. An opinion after three months covers things an unboxing cannot, such as battery wear.
  • Specific details. They name the model, a setting, a size or a measurement. “The sizing kit runs small” tells you more than “bad fit”.
  • Use in context. They say where, when or with what. “On night shifts” or “with a sensitive scalp” ties the verdict to a routine.
  • Mixed verdicts. They praise one feature and criticize another.
  • Follow-ups. They post an update after some weeks, or answer another person’s question with detail.
  • The product in use. In a video, they wear, apply or operate the product, not just unbox it.

No single sign is proof. Someone can write “I bought it” without buying it. But several signs together make an opinion stronger evidence. You can also check specific details against what other owners say.

Here is how five opinions about one smart ring compare.

OpinionWhat it showsHow to weigh it
“Best ring ever, buy it.”No sign of useWeak: anyone could write it
“Two months in, mine lasts four days, not seven.”Says they used it, with detailStrong
“My sister says the battery dies fast.”Passes on someone else’s experienceUseful, but report it apart
“The brand says it lasts a week.”Repeats the brand’s claimNot an owner’s opinion
“Does the new model last longer?”A question, not an opinionShows what a shopper doubts

What makes an opinion weaker evidence?

An opinion is weaker evidence when nothing in it shows use. That does not make it untrue. It means it supports a different kind of finding.

Watch for these patterns:

  • Generic praise or dislike. “Love it” or “Avoid” with no feature and no detail.
  • Someone else’s experience. “My friend says it broke.” Report these apart from people who say they used it.
  • The brand’s claims. The text repeats the spec sheet, the ad or the brand’s own words.
  • Opinions before launch. Few people can use a product before it goes on sale. Early posts show hopes and doubts, not use.
  • Copied text. Many accounts post the same words in the same days. Count identical text once.
  • Questions. “Does it run small?” is not an opinion, but it shows what stops a purchase.

Keep weaker evidence, label it, and read it for what it can tell you. Doubts from people who have not bought yet are a strong guide for your product page and your ads. Deleting them hides the objections a marketing team most needs to answer.

Does a sponsored review mean the person did not use it?

No. A sponsored post is one where the brand paid the creator or sent the product. Many sponsored creators use the product in their videos. The problem is the incentive, not a lack of use.

Signs that a post is sponsored:

  • a platform label for a paid partnership;
  • a tag such as #ad, #sponsored or #gifted;
  • a line saying the brand sent the product;
  • a discount code or an affiliate link, which shows a business tie but not always payment.

Store reviews have their own version. Some stores mark reviews written in exchange for a free product. Read those apart from reviews by people who paid.

Some deals are never disclosed, and a deal that nobody disclosed does not show in the post. So label the sponsorship you can detect, and read sponsored and organic opinions apart. If both groups say the same thing, the finding is stronger. If they differ, report both. The 30-day launch read shows how to do this week by week.

How do you check opinions at scale, step by step?

Reading one review is a judgment call. Reading thousands needs rules that you apply the same way each time. Use this checklist.

  1. Set the scope. Name the products, markets and time window before you collect anything.
  2. Label the basis. Use four labels: says they used it, passes on someone else’s experience, repeats the brand’s claim, or does not say.
  3. Mark how specific it is. Use vague, specific or detailed. A detailed opinion from someone who says they used it carries the most weight.
  4. Label sponsorship. Mark the posts that carry a disclosure. Keep them in the data and read them apart.
  5. Count identical text once. The same words from different accounts can make one view look common.
  6. Count accounts, not posts. One person who posts ten times is still one person. One person with two accounts still counts twice, so say “accounts”.
  7. Compare streams. Store reviews, captions, what is spoken in videos and comments differ. Count each stream apart, and compare like with like.
  8. Read a sample by hand. Check that each label fits the words. If many do not fit, fix the rules and label again.
  9. Name the base for every number. Write “X of N opinions from people who say they used it, on P platforms, from month to month.”

Do not skip step 8. A label is a reading of words, and words can mislead. Sarcasm, slang and other languages cause errors for people and for software. A sample read by hand shows how far you can trust the labels.

Step 9 keeps your readers honest too. A number with no base invites the wrong comparison. “Most owners like the battery” says less than “X of N opinions from people who say they used it praised the battery”.

What can these signs not tell you?

They tell you what a person says about their own use. They do not prove a purchase, and they do not show what silent buyers think.

  • A statement of use is not proof. A person can claim use they never had. Treat the label as “says they used it”, never as more.
  • Silent buyers stay silent. Many buyers never post. The opinions you read come from people who chose to speak.
  • Counts are not shares of buyers. “X of N opinions” describes your collection, not the market.
  • Writing style is a weak sign. Polished text can come from a careful owner or from software. Judge the details, not the style.
  • Coverage is uneven. A new or small product may have few opinions from people who say they used it. Say so when the base is thin.

When you need a share of all buyers, field a survey. Use public opinions to find the right questions and the words people use. The guide to preliminary consumer research shows how to do that.

Where does Honestly fit?

Honestly is a proprietary dataset of product opinions. Most sources are public; the record of each opinion, linked to its source, is ours. The data streams include videos and what is said in them, posts, comments, replies, reviews and forums.

Honestly labels opinions where the person says they used the product. The label is a model’s reading of the words, not proof of use or purchase. It also misses some people who do say it. Honestly labels the posts it detects as sponsored.

Opinions without a sign of use stay in the dataset. One platform serves consumer insights and marketing teams. Insights teams can read what people who say they used it praise and criticize. Marketing teams can read the doubts of people still deciding. Then go deeper with people who say they used the product.

Our published studies are on the research page.

Frequently asked questions

How can I tell if a review is real?

You cannot prove it from the words alone. Look for words of own use, specific details and a mix of praise and complaints. An opinion without those signs is weaker evidence, but it is not necessarily untrue.

Can you tell if a review was written by AI?

Not reliably from the writing style. Polished text can come from a careful owner, and clumsy text can come from software. Judge the details instead, and check them against what other owners say.

How do I tell if a product review is sponsored?

Look for a paid-partnership label, a tag such as #ad, or a line saying the brand sent the product. Discount codes and affiliate links show a business tie. A deal that nobody disclosed does not show in the post.

Should I delete opinions that do not say the person used it?

No. An opinion that does not say so is not proof that the person did not use it. Label these opinions and read them apart. Questions and doubts from shoppers show what stops a purchase, and your marketing team needs to know that.

What does “people who say they used it” mean?

It means the person states in their own words that they used, owned, bought or tested the product. It is a reading of what they wrote or said, not proof of purchase. We use the phrase because it claims no more than the words show.

See which opinions about your product come from people who say they used it.

Request a demo