Web Design
Landing Pages & Funnels
8 Ways to Build Trust on a Website: Complete 2026 Guide

Joshua Jackai
17 min

Published September 29, 2026 ยท Last updated and fact-checked October 31, 2026
Short answer: Build trust on a website by putting named, dated proof beside every call to action, not on one testimonials page. In a Stanford study of 2,684 people, a site's look drew 46.1 percent of credibility comments (Fogg et al., 2003). Then measure each ask separately and keep every claim checkable.
Key takeaways
A site's visual look drew 46.1 percent of the comments about credibility in a study of 2,684 people (Fogg et al., 2003).
Online book reviews changed sales, and one-star reviews moved them more than five-star ones (Chevalier and Mayzlin, 2006).
People formed a stable opinion of a page's appeal in 50 milliseconds (Lindgaard et al., 2006).
About 35 percent of pages published since ChatGPT showed signs of AI authorship, against roughly 10 percent of the whole web (Pew Research Center, 2026).
My W Communications rebuild carried proof on every page instead of one testimonials page; it reports what was built, not what it produced.
Way | Best for | Effort | Key number |
|---|---|---|---|
1. Put proof beside the ask | Pages that ask for an action | One afternoon per page | 1 call to action per page |
2. Retire the testimonials page | Sites with proof in one room | A few hours | 46.1 percent |
3. Name and date every proof point | Case studies and claims | Ongoing | 3 required details |
4. Cut the distrust signals | Pages with badges and timers | Under an hour | 5 faults |
5. Win the first glance | Landing pages | A design pass | 50 ms |
6. Measure each ask separately | Pages with two or more buttons | One tagging pass | 2 buttons, 2 tags |
7. Make claims checkable for AI | Sites cited by answer engines | A profile audit | 35 percent |
8. Trust with no reviews yet | New sites and new studios | One detailed write-up | 0 reviews needed |
I build websites for a living, so I have an interest here. Demand Designer is a fixed-price web design studio for B2B companies, run by me, Joshua Jackai. This guide is for marketing leads and founders who want a visitor to believe the page before they act on it. The worked example is W Communications, LLC, a client project of the studio, built between May 2025 and June 2026.
1. Put proof beside the ask: best for pages that ask for an action
Put proof beside the call to action, not one or two clicks away from it. A call to action is the single thing a page asks a visitor to do. Every conversion path has a moment where the visitor has to believe something, and that is where the evidence belongs.
What it is: a named client, a dated project or a full-name quote on the same screen as the ask
Where: every page with a call to action, not only the homepage
Effort: about an afternoon per page
Rule I use: one call to action per page
A conversion path is the route from landing on a page to completing that one action. My rule is one call to action per page, because a second ask makes a visitor choose between the two before choosing the one you wanted. That is a design position, not a finding.
"Proof belongs beside the ask. Not near it. Beside it.", wrote Joshua Jackai, founder of Demand Designer, on where evidence goes.
The research gives a reason to care about the moment of belief. McKnight, Choudhury and Kacmar wrote that trust plays a central role in helping consumers overcome perceptions of risk and insecurity in e-commerce. If the proof for that belief sits two clicks away, the evidence is furthest from the ask at the point that carries the most weight. I treat placement as a design decision you make page by page. In the Stanford credibility study of 2,684 people, the look and structure of a site dominated the comments, so a quote that is hard to see on a crowded screen does little. Make the proof as easy to spot as the button it supports, and keep it to one or two items, not a wall of logos that competes with the ask.

2. Retire the testimonials page as the home for proof: best for sites that keep proof in one room
A testimonials page can only be read by someone who clicked through to it. That reaches people already interested enough to go looking. Keep the page if you like, but spread its content across the pages that ask for something, so the least convinced reader still meets the evidence.
The problem: proof sits behind a door the least convinced reader has little reason to open
The fix: move quotes, logos and case studies onto the pages that ask
What stays: the testimonials page, as an archive
My caveat: I have no data on how testimonials pages perform
My argument is that placement matters more than volume. Gathering more quotes, more logos and more case studies leaves all of it in one room. The same material, distributed across the pages that do the asking, is present in every path instead of parked in one. I am not claiming a measured lift for this, and you should test it on your own pages. A simple way to test it is to move one quote beside the main button on your most visited page, wait a full sales cycle, and compare enquiries per visit against the previous cycle.
Reviews do change behaviour when people see them. Chevalier and Mayzlin studied book reviews at two retailers and found that an improvement in a book's reviews leads to an increase in relative sales, and that for most samples the impact of one-star reviews was greater than the impact of five-star reviews. Readers also seemed to read review text, not only the star average.
"the impact of one-star reviews is greater than the impact of five-star reviews", wrote Chevalier and Mayzlin in the abstract of their 2006 study.
That finding argues for showing real, specific proof, including the awkward kind, where the reader will weigh it.
3. Name and date every proof point: best for case studies and claims
Name the client, date the work and state the constraint. A case study with no timeframe, an award with no year or a quote with no full name asks the reader to guess. Specific, dated proof can be checked against something outside the page, and that is what makes it believable.
Name: a real client or person, with a live URL where you can
Date: a month and year for the work, the award or the quote
Constraint: the limit you worked inside, such as a launch date that could not move
Scope: say what the proof covers and what it does not
W Communications, LLC is a strategic communications consultancy in Seattle, four decades old, founded by Wanda J. Herndon, who previously led global communications at Starbucks. The live site is wcommgroup.com. Demand Designer rebuilt it as a client project, and the engagement ran from May 2025 to June 2026. The build moved eight pages of hand-coded HTML into WordPress on Bricks Builder, a visual page builder that lets a non-technical team edit a page without touching code. It organized ten service disciplines, three conversion paths and one call to action per page.
The delivery was phased because an event date could not move. A single landing page for the London book launch event shipped ahead of the full rebuild. I wrote up that side separately, in how an executive website gets built and handed over to a non-technical team. What matters here is where the proof went: case studies, testimonials and endorsements were carried across every page rather than parked on one.
Wanda J. Herndon wrote a book, Working Class to Breaking Glass: One Woman's Fight to Belong and to Lead, published by Ingenium Books. Howard Schultz, Christine Day and Carolyn House Stewart endorsed the book, not the website, and the site carries the endorsements as what they are. Stretching a book endorsement into a website endorsement would have been an easy sentence to write, and a bad trade. This guide reports what was built and the reasoning behind it, not what it produced.
"Stretching a book endorsement into a website endorsement would have been an easy sentence to write, and a bad trade", wrote Joshua Jackai, founder of Demand Designer, on the W Communications endorsements.

4. Cut the distrust signals: best for pages crowded with badges and timers
A few faults reliably damage credibility, and removing them is quicker than adding proof. These are my judgments as a reader, not measured effects. Stacked badges, undated work, unsourced claims, fake urgency and broken links each make a visitor wonder what else is off.
Stacked badges: five security seals and four award logos in a row read as anxiety
Undated work: a post, award or case study with no date
Unsourced claims: "studies show" with no institution, date or sample size
Fake urgency: a countdown timer that resets when the page reloads
Broken links: a buy button for the wrong edition, a form that sends nowhere
One mark a reader can verify is worth more than nine that decorate. A claim needs the institution, the date and the sample size, and a deadline needs to be a real one. Anyone who reloads the page can see a timer reset, and I would not put one on a client's site. Faults like a wrong link are easy to miss in a desktop review, and they surface at the moment someone is trying to act.
In a published study, Fogg and colleagues asked 2,684 people to judge the credibility of live sites. The reasons people gave were varied, and the study found that design look was mentioned in 46.1 percent of comments, with information structure and focus next. That supports treating polish and clarity as part of trust, though it does not rank individual badges or timers. In that study, 2,684 participants wrote about live sites, so the finding rests on what people noticed, not on what they bought. I would still remove a fault I can see, because it costs nothing to fix.
"One mark a reader can verify is worth more than nine that decorate", wrote Joshua Jackai, founder of Demand Designer, on stacked badges.

5. Win the first glance: best for landing pages and paid traffic
Visitors judge a page almost before they read it. Lindgaard and colleagues found that people formed a reliable view of visual appeal in 50 milliseconds, and the quick ratings matched longer ones. For a landing page that means layout, type and clarity do trust work before any proof is read.
Window: about 50 milliseconds for a first visual impression
What counts: layout, type, spacing and image quality
What does not rescue it: proof placed below a cluttered first screen
Check: open the page on a phone and look at it for one second
The two studies fit together. Lindgaard's team measured the speed of the judgment, and Fogg's team found that the look of a site was the most common thing people mentioned when explaining credibility. I read them as one point: the first screen has to look careful, and then the proof has to be visible on it. The 50 millisecond result came from a lab task on rated screenshots, so I treat it as a reason to design the first screen with care, not as a rule about any one pixel. Check your own first screen on a phone and ask what a stranger would conclude. A careful first screen with no proof is polite. Proof on a messy first screen is invisible.
"Attention web designers: You have 50 milliseconds to make a good first impression!", wrote Lindgaard and colleagues in the title of their 2006 paper.
This is also why stock photos work against you. A visitor who has seen the same photo on ten other sites reads yours as one more of them. Use your own screenshots, your own people and your own work wherever you can.

6. Measure each ask separately: best for pages with more than one button
You cannot measure belief directly, but you can measure the action that follows it. Measure each ask on its own, and report one line per button, not a single combined number for the site. Otherwise you cannot tell which piece of proof sat next to which action.
Method: one tag for each button and each link
Tool: a UTM parameter, a tag added to a link so analytics can report its source
Catch: Amazon strips UTM parameters on its own servers
Fix: capture the outbound click on your own site before the visitor leaves
Attribution is the practice of working out which link, page or campaign a visit or a click came from. On the W Communications books page, two Amazon buy buttons sit on one page, and in my experience two buttons on one page can report as one line in analytics unless each carries its own tag. Per-button tagging was added so each button reports separately.
One catch matters before any of that gets built. Amazon strips UTM parameters on its servers, so a tag on an outbound link to Amazon does not survive the trip. Attribution for an outbound click has to be captured on your own site, at the moment of the click. A measurement plan that depends on the destination honoring your tags is not a plan. The tagging exists so the question can be asked later with real numbers, and this guide does not report those numbers. If you run a similar page, write down the question before the first visit arrives, such as which button gets more clicks, so the data answers something you already asked.
"A measurement plan that depends on the destination honoring your tags is not a plan", wrote Joshua Jackai, founder of Demand Designer, on outbound links.

7. Make claims checkable for AI answer engines: best for sites that want citations
An answer engine answers questions instead of listing links, so a website is now read twice: by a person and by a model deciding whether to cite it. Fluent copy no longer proves anything. What remains is what can be checked outside the page.
Person: a named author in a role that can be verified on a profile
Client: a named client with a live URL
Time: a dated engagement
Consistency: claims that match third-party mentions of you
The Pew Research Center published an August 2026 report titled "How Much of the Internet Is Written With AI?", based on about 490,000 English-language pages from Common Crawl, a public archive of crawled web pages. Pew found that 35 percent of pages published since the release of ChatGPT show signs of AI authorship, against about 10 percent across the whole web. In a July 2026 sample, .com pages sat near 9.4 percent while .edu and .gov pages sat near 1 percent. Pew warns that detection is imperfect.
"significant signs of AI authorship", wrote Pew Research Center in its 2026 report, defining what its detector counted.
My reading is that fluent, reassuring copy is now close to free, so fluency has stopped being evidence. Entity consistency is the working term for what is left. An entity is the thing a page is about: a person, a company, a product. Consistency means the name, role, location and claims agree across your own pages, your profiles and third-party mentions. I work on the assumption that a site contradicting its own founder's public profile gives a model a reason to pick a different source. That is my inference, not a measured result.

8. Build trust with no reviews yet: best for new sites and new studios
A new website has no client roster and no review count, so specificity has to do the work that volume normally does. Describe your process in detail, put a real name and face on the page, date everything and describe one piece of work so precisely that only someone who did it could have written it.
Process: list the steps, including the parts you refuse to do
Scope drivers: page count, service areas, an existing site to move and any date that cannot move
Person: a real name and face in a role that can be checked elsewhere
Proof: one job in operational detail, with its constraint, not the outcome you wish you had
A reader who can see what changes the shape of the work can also see that you have done it before. Name the constraint you worked inside rather than the result you wish you could claim. Then answer, in your own words, the objection you are most afraid of. A buyer is already thinking it, and a page that says it first reads as confident. The Stanford study found that information structure and focus were the next most common things people noticed after look, which is exactly what a clear, detailed write-up supplies (Fogg et al., 2003).
Start with the page you would least want a stranger to land on. Open your site in a private browser window on a phone and go to the page a stranger is most likely to reach from a search result. In my experience it is rarely the homepage. Find the one action it asks for, then look at what proof is visible on the screen at the moment it asks.
If nothing is there, move one piece of proof onto that page this week: a named client, a dated piece of work, or a quote with a full name and title. One page, one path, one piece of proof, then the next page, until every route a buyer can take has evidence inside it. The testimonials page can stay. It just stops being the plan.
"The testimonials page can stay. It just stops being the plan", wrote Joshua Jackai, founder of Demand Designer, on where proof goes.
How the eight ways compare: filter by effort and job
Use this matrix to pick where to start. It lists each way with its job, effort and the evidence behind it, and the filter hides rows without deleting them. The static table is the same data, so the page reads fully without scripts.
Way | Job | Effort | Evidence type |
|---|---|---|---|
1. Proof beside the ask | Conversion | Medium | Design position, McKnight et al. 2002 |
2. Retire the testimonials page | Conversion | Low | Chevalier and Mayzlin 2006 |
3. Name and date proof | Credibility | Medium | Design position |
4. Cut distrust signals | Credibility | Low | Fogg et al. 2003 |
5. Win the first glance | First impression | Medium | Lindgaard et al. 2006 |
6. Measure each ask | Measurement | Low | First-party experience |
7. Checkable for AI | AI visibility | Medium | Pew Research Center 2026 |
8. Trust with no reviews | New sites | Medium | Design position |
How many pages need proof? A quick page audit
Count the pages where a visitor can take an action and the pages that already show named, dated proof beside it. The gap between the two is your to-do list. Enter your own numbers in the interactive version of this page; the defaults are placeholders, not benchmarks.
For example, if a site has 12 pages with a call to action and 3 of them show proof beside it, 9 pages ask without proving. At one afternoon per page that is roughly nine afternoons. Start with the page most likely to be reached from search, as in way 8, then work down the list. The arithmetic is simple on purpose: nothing here is a sourced constant.
Frequently asked questions
Should testimonials go on their own page?
A testimonials page can exist, but it should not be the plan. Reaching it requires a visitor who already went looking, so put case studies, testimonials and endorsements beside the call to action on every page instead, where the least convinced reader will see them.
How do you measure whether a website is building trust?
Measure each ask separately rather than reporting one combined number, so every button and conversion path has its own tag. Capture outbound clicks on your own site, because Amazon strips UTM parameters on its servers before the click is ever reported back to you.
How do AI answer engines decide a website is credible?
They can check specifics against sources outside the page: a named person in a verifiable role, a named client with a live URL, a dated engagement and claims that match third-party mentions. Pew found 35 percent of post-ChatGPT pages show signs of AI authorship, so fluent copy alone proves little.
How do you build trust on a new website with no reviews yet?
Specificity replaces volume. Describe your process including what you refuse to do, name what drives the size of a job, put a real name and face in a checkable role, date everything, and write up one job in enough detail that only someone who did it could have written it.
Want proof sitting next to every ask on your site?
Demand Designer rebuilds B2B and executive sites so the evidence lives inside the conversion path instead of behind a testimonials link. If you want to find where your own site asks without proving, the Pipeline Leak Scorecard scores it in about fifteen minutes, or book a strategy call and I will walk your pages with you.
Sources
Peer reviewed
Fogg, B. J., Soohoo, C., Danielson, D., Marable, L., Stanford, J., and Tauber, E. R. (2003). How do users evaluate the credibility of Web sites? Proceedings of the 2003 Conference on Designing for User Experience. https://doi.org/10.1145/997078.997097
McKnight, D. H., Choudhury, V., and Kacmar, C. (2002). Developing and validating trust measures for e-commerce: an integrative typology. Information Systems Research. https://doi.org/10.1287/isre.13.3.334.81
Chevalier, J. A., and Mayzlin, D. (2006). The effect of word of mouth on sales: online book reviews. Journal of Marketing Research. https://doi.org/10.1509/jmkr.43.3.345
Lindgaard, G., Fernandes, G., Dudek, C., and Brown, J. (2006). Attention web designers: you have 50 milliseconds to make a good first impression! Behaviour and Information Technology. https://doi.org/10.1080/01449290500330448
Industry and vendor (not peer reviewed)
Pew Research Center (August 20, 2026). How much of the internet is written with AI? https://www.pewresearch.org/data-labs/2026/08/20/how-much-of-the-internet-is-written-with-ai/
W Communications, LLC, live site, checked October 2026. https://wcommgroup.com
Free workbook
Pipeline Leak Scorecard
Find the leaks between your traffic and your pipeline. Score your site in about fifteen minutes, with it open in front of you.
12 questions, 8 pages, 15 minutes. Free.







