Co-design means the people affected by a design hold power over it — from the beginning, not at the end.
That is the whole idea. Everything difficult about co-design is in the gap between that sentence and what gets called co-design in practice.
Co-design can be defined as the process of active collaboration between stakeholders in designing solutions to a prespecified problem.
We understand co-design as a process of joint inquiry and imagination, as stated by Steen, where stakeholders are empowered to participate in inquiry (explore, discuss, and define a problem) and imagination (explore, develop, and evaluate possible solutions) through a series of interconnected and iterative phases.
Note what “joint inquiry” includes: defining the problem. If the problem arrives already defined and you are asked to help solve it, that is not co-design. That is the part people skip.
Our position
Co-design isn’t consultation at the testing stage. It is power-sharing at the architecture stage. Those are different things.
“Nothing about us without us” is not a design methodology. It is a political demand.
Inviting feedback is not the same as sharing power. Inclusion that stops at “asking for opinions” misses control, resources, and influence.
Co-design is the operational form of nothing about us without us. It is what the demand looks like when it reaches a process, a budget, and a decision.
Co-design that isn’t intersectional reproduces the exclusion it claims to fix — earlier in the pipeline, where it is harder to see. A process that doesn’t include Disabled people of color, multiply-marginalized Disabled people, and people whose disabilities are least legible to mainstream accessibility frameworks moves the exclusion from deployment back to design. That is not progress. It is relocation.
The demand comes from us
Co-design is not an accessibility researcher’s recommendation handed down to Disabled people. It is what Disabled people say they want, when asked.
A 2026 poll of over 1,000 disabled UK adults — conducted by Business Disability Forum with Opinium — found that 40% identified co-design as the top requirement for making AI more accessible: designing, developing, and testing AI products with disabled people from the outset. That’s not an accessibility researcher’s recommendation. That’s disabled people naming the structural gap in their own words.
Ranked first. Above every other fix on the list.
The word is claimed far more often than it is practised
This is measurable, and it has been measured.
A 2025 systematic review in the Journal of Public Health examined 55 published studies that reported using co-design. What it found should be read next to every project that describes itself with the word.
Forty-four per cent never defined co-design at all.
Forty-four per cent (n = 24) of selected articles did not define co-design.
Participation was mostly low.
The extent of participant engagement ranged from low to low-medium, across most co-design steps, depending on the level of agency, input, and empowerment afforded by the methods.
Seven per cent did anything about power.
For example, only 4 (7%) included studies reported strategies to mitigate power dynamics.
And the review names the failure mode in plain words: extraction.
This review highlights a significant issue in co-design practices within public health: the variability in stakeholder participation, often criticized for being extractive rather than generative. This extractive approach gathers information from stakeholders without fostering their ownership and engagement through active collaboration.
Extractive rather than generative. That is the distinction to carry around. It is not about whether people were in the room. It is about who left the room owning something.
The ladder: which role did people actually get?
The review classified participants using the MATE taxonomy of participatory design, which sorts the roles a participant can hold at each stage. It is the most useful diagnostic on this page, because it replaces a yes/no question with a specific one.
The six roles: informant, consultant, co-designer, tester, leader, trainee.
Engagement is scored on three factors — agency, input, and empowerment.
Across the 55 studies, participants were mostly informants and consultants. Few projects reached the point where participants tested a prototype. Almost none reached leader.
So the question to ask of any process calling itself co-design is not “were Disabled people involved?” It is:
- Which role did they hold, at which stage?
- Did they help define the problem, or only respond to one?
- Who made the final call when there was disagreement?
- Were they paid?
- Do they own any of the output?
An informant is not a co-designer. Both can be described as “involved.”
Design thanking: what extraction feels like from inside
Liz Jackson named the emotional shape of this years before the systematic reviews caught up.
Step one of design thinking is cultivating empathy through observations and interviews, but to a disabled person that can feel a little bit less like empathy and a little bit more like designers gleaning our ideas and our life hacks and they’re selling them back to us as inspirational do good.
Step two is defining the problem, but because disabled people are unable to lead these processes, it is us that’s defined as the problem rather than the problem being defined as the problem.
— Future is Code for Eugenics | Liz Jackson | 2019 Core77 Conference
So you have our insights gleaned, we are defined as a problem, and then designers enter this iterative process of ideation prototyping and testing which I argue leads to the unacknowledged sixth-step of design thinking, or as I call it design thanking, because we’re expected to be grateful for that which has been done for us.
— Future is Code for Eugenics | Liz Jackson | 2019 Core77 Conference
Design thanking is what co-design becomes when the word survives and the power-sharing doesn’t. Our ideas go in. Someone else’s product comes out. We are expected to say thank you.
The output of that process has a name too: the disability dongle.
The language gives it away
Bryony Shannon’s audit of co-production vocabulary is the fastest way to tell whether an organization has understood the word or just adopted it. Her point is that the term itself is fine — it’s the words that cluster around it that expose what’s really happening.
But once we explore the language associated with co-production, it quickly becomes clear that while the meaning of the term couldn’t be more explicit, we’ve largely managed to miss the point in our interpretation.
On “professionals and service users”:
So, we’ve already failed. Instead of working together as equals, our basic definitions imply and perpetuate difference and inequality. Us and them.
On “engagement”:
‘Engagement’ is something we do to people, at a fixed time, for a fixed purpose. There’s no sense of an evolving, organic, long-term relationship, where trust is built and where genuine, human connections are made.
And if you choose not to show up when we tell you to, we’re quick to slap on a ‘refuses to engage’ label and move on.
On “empowerment”:
We – the ‘professionals’ – give you – the ‘users’ – power. If we feel like it. Maybe. Sometimes. Or maybe not.
We see power as ours to give away.
And, by default, to take away.
“Empowerment: An illusory gift from professional to a not quite human.”
— Mark Neary, quoted in Co-production: words that make me go hmmm…
On “you said, we did”:
You told us some things and we went away and did stuff without you.
Erm – so much for co-production.
And the line that decides whether any of it is real:
But co-production isn’t extra work, it is the work.
A co-design lead, a co-design strategy, and a half-day training course are not co-design. They are a place to put co-design so that it doesn’t disturb anything.
Participation can be a decoy
There is a structural risk here that disability justice requires us to name, and it is not the same as tokenism.
Tokenism is co-design done badly. A decoy is co-design working exactly as intended — for someone else.
Ironically, the inevitability decoy extends to efforts to engage on-the-ground stakeholders through participatory design… Such work therefore feeds rather than resists the Project of AI.
— Reckoning with the Political Economy of AI: Avoiding Decoys in Pursuit of Accountability
Participation can be conscripted. When the only question on the table is how should this be built, agreeing to answer it concedes whether it should be built, and who decides. Community voice then becomes the thing that makes a predetermined outcome look consented-to.
The decoy is participation that launders the Project’s logic through the appearance of community voice. The demand is structural power over what gets built and whether it gets built at all.
This is not an argument against community involvement. It is an argument for keeping “should this exist?” inside the scope of the co-design, and for treating a process that has ruled that question out of order as something other than co-design.
Partial co-design is a real thing — say so
Co-design is a gradient, not a binary, and the honest thing is to name where on the gradient you actually landed.
Fisher and colleagues did exactly that. They describe their own work as a partial co-design framework, and then say precisely what was missing:
Of note, the current study took only a partial co-creation approach, as autistic adults were not formally consulted in the conceptualization of the research program (beyond the anecdotal feedback described above), prior to the finalization of the study design. To align with full best practice co-creation principles, autistic adults should be formally engaged at the outset and involved in planning, exploration, development, decision points, and change/revisions aspects of the project.
— Developing a neurodiversity-affirming group therapy program for autistic adults through co-design
Four participants, giving feedback on a program that was already drafted. That is consultation, and they call it consultation. Naming the limit is what makes the paper trustworthy — and it hands the rest of us a usable sentence: engaged at the outset, and involved in planning, exploration, development, decision points, and revisions. That is the checklist. Anything less is partial, and partial is fine as long as nobody calls it full.
There is a second honest caution, and it cuts the other way — toward processes that look mature.
However, at the same time, researchers should guard against working with a selected group of community representatives over a very long period, where all parties become bought-in to a shared agenda. One function of inclusive research is for team members to be confronted with perspectives other than their own, and this value may dissipate over a long partnership which becomes an echo chamber.
A comfortable long-running co-design group can stop being a check on power and start being a source of agreement. Both failure modes are real: too little community power, and a community presence that has been absorbed.
What it actually takes
Pay people
This is the shortest test, and the one most processes fail.
One way to promote empowerment for community partners is to provide adequate payment for their time, at consultancy (rather than research participant) rates. Meetings at which only some people are paid to be there are not best inclusive practice. Payment explicitly values expertise by lived experience, and means that joining a research team is more likely to be framed as a right, rather than as a privilege.
Meetings at which only some people are paid to be there are not best inclusive practice. Read that as a room test. Look around the meeting and count who is being paid to be there. If the Disabled people are the unpaid ones, the power arrangement is already visible.
The paper also does the unglamorous work of naming why payment fails in practice — welfare rules that make consultancy income costly to accept, and institutional finance systems that cannot pay a community expert — and treats those as problems for the organization to solve, not reasons to stop paying.
Share the authorship
We highly recommend collaborative analysis, interpretation and co-authorship of project outputs as a valuable way to reduce power disparities and deliver equality.
The NEST toolkit shows what this looks like on the page. Its co-designers are credited by name, in two groups — Neurodivergent young people and Neurodivergent adults — including the ones who chose not to be named:
Co-design teams: Our young person co-design team included Cameron Coupar, Jamie Yates, Minerva Fletcher-Watson, Jacob McKinlay and four others who chose not to include their names. The adult co-design group included Carrie Watts, Dr Elliott Spaeth, Harriet Axbey, Marisa McKinlay, Natalie Macpherson, Nicola Daniel, Rose Bangs, Victor, and one other who chose not to include their name.
And the handbook records where co-designers overruled the default. The inclusion criteria are broad — diagnosed, waiting for diagnosis, self-identifying, or wondering — because the co-designers wanted it that way:
This was a strong preference of our co-designers and fits with the inclusive ethos of the NEST group.
That sentence is evidence. It shows a decision that went the co-designers’ way and is documented as theirs. A co-design process that produced no such sentence probably didn’t share any decisions.
Make the roles paid, permanent, and senior
England’s NHS culture of care standards set the bar for mental health inpatient services, and they draw the consultation line explicitly:
Integrating lived experience takes the form of true co-production, not just consultation or engagement. All training is co-designed and co-delivered by people with lived experience.
— Culture of care standards for mental health inpatient services
Alongside paid lived experience and peer roles, the infrastructure supports a broad range of voices to be heard, and prevents lived experience becoming assimilated or co-opted.
— Culture of care standards for mental health inpatient services
We are committed to growing lived experience leadership roles across the organisation, including a patient director and patient safety partners, to ensure co-production across service design, quality improvement and oversight. Those in lived experience roles can access training and development, and have a clear career pathway, alongside opportunities to move into different roles. People with lived experience are employed at a range of bandings.
— Culture of care standards for mental health inpatient services
A career pathway. Bandings. Someone in the room whose job is lived experience leadership and who can be promoted. That is co-design with a budget line, which is the only kind that survives a reorganization.
Hold the principles
Design Justice states the standard more compactly than anyone:
- We see the role of the designer as a facilitator rather than an expert.
- We believe that everyone is an expert based on their own lived experience, and that we all have unique and brilliant contributions to bring to a design process.
- We work towards sustainable, community-led and -controlled outcomes.
Community-led and community-controlled. Not community-consulted.
The short version
Co-design is a claim about power, so audit it like one.
- Who defined the problem?
- Who is paid to be in the room?
- Which role did we hold — informant, consultant, co-designer, tester, leader?
- Could the answer have been “don’t build this”?
- Whose name is on the output?
- What decision went our way, and where is it written down?
If the answers are uncomfortable, the process may still be worth doing. Just don’t call it co-design.