What accommodations can you ask for on a timed coding assessment if you have dyslexia?

The ones worth requesting are a time multiplier applied to the invite before it is sent, confirmation that a read-aloud or text to speech tool is permitted, the ability to change font and theme in the editor pane, a single overall time budget instead of a per-question countdown, confirmation that spelling is not scored in written sections, and an untimed practice question on the same platform beforehand.

Do you have to disclose dyslexia to get a coding interview accommodation?

Usually not for a format change. Describe the barrier and the adjustment instead of the diagnosis: "I read long prompts more accurately with more time, so could the assessment be issued with extra time?" is a complete request. Employers can sometimes ask for supporting information, and the rules differ by country, employer and role, so treat this as wording help rather than legal advice.

How do you make a dense LeetCode problem readable with dyslexia?

Read the return sentence and the constraints block first, then rewrite the whole prompt as three bullets: what you are given, what rule you apply, what you hand back. Then build one small example input of your own that hits the clause the prose buried, run it by hand, and work from your restatement instead of re-reading the paragraph.

Can you use text to speech during an online coding assessment?

Often yes with permission, but ask two questions rather than one. Permission from the recruiter is the easy half. The second half is whether the proctoring configuration allows it, because a lockdown browser or a screen recording setting can block other applications regardless of what the recruiter agreed. Get both answers in writing before the invite is generated.

How much extra time do candidates get on an online assessment?

There is no single standard. Assessment platforms commonly expose a per-candidate time multiplier that the employer sets, and requests are often phrased as 25 percent or 50 percent more time, but what is granted depends on the employer and the platform. Ask for a specific multiplier rather than "extra time", and ask before the invite is generated.

When should you request a dyslexia accommodation for a coding interview?

At the moment you are being scheduled, and before the assessment invite is generated. Send it to the recruiting coordinator who owns the invite rather than the hiring manager, and ask whether there is a dedicated accommodations contact. Timer changes are the ones that go stale fastest, because on many platforms the setting is attached to the invite itself.

Dyslexia Coding Interviews: Cutting the Reading Bottleneck

The bottleneck in a coding interview is almost never the algorithm. It is the 150 words of indexed prose in front of it, read under a countdown. Two things fix most of that: a restating routine you run on every prompt before writing a line of code, and a time multiplier applied to the assessment invite before it is generated. Below is the routine worked end to end on a deliberately dense prompt, then eight accommodations worth requesting on a timed online assessment with the sentence to send for each.

This is wording and technique, not legal advice. In the United States these are reasonable accommodations under the ADA, in the UK they are reasonable adjustments under the Equality Act 2010, and both are framed around the barrier rather than the label. What applies to you depends on where you are, the employer, and the role.

Worked example: turning a dense prompt into three bullets

Here is the kind of paragraph that eats eight minutes of a sixty minute assessment. Read it once. Do not try to hold it.

You are given a 0-indexed 2D integer array requests where requests[i] = [from_i, to_i, time_i] indicates that user from_i sent a follow request to user to_i at minute time_i, an integer array limits of length n where limits[u] is the maximum number of follow requests user u will accept within any window of k consecutive minutes, and an integer k. Requests are processed in the order in which they are given. A request is accepted if and only if accepting it would not cause the recipient to exceed limits[to_i] within the window of k consecutive minutes ending at time_i inclusive; otherwise the request is silently dropped and does not count towards any future window. Return the total number of accepted requests. If requests is empty, return -1. Constraints: 1 <= n <= 10^5, 1 <= k <= 10^9, and time_i is strictly increasing.

Now the same problem as three bullets. This is what you type into the comment block at the top of your editor, in your own words, before touching the solution.

What I get. A list of requests already in time order. Each one is three numbers: sender, receiver, minute. Plus a cap per receiver, and a window length k.
What I do. Walk the list once. Accept a request if the receiver has fewer than their cap of already accepted requests inside the last k minutes, counting backwards from this request's own minute. Dropped ones never count.
What I hand back. The number of accepted requests. Empty input is the trap: it returns -1, not 0.

Then one example input, small enough to run in your head, chosen so it hits the clause the prose buried rather than the one the prompt already illustrated.

requests = [[0,1,1],[2,1,2],[3,1,9]], limits = [5,2,5,5], k = 5
Minute 1: user 1 has nothing in the window. Accept. User 1 now holds one accept, at minute 1.
Minute 2: the window covers minutes -2 to 2 and holds one accept. The cap is 2, and 1 is under 2. Accept.
Minute 9: the window covers minutes 5 to 9. Both earlier accepts have fallen out of it. Accept. Answer: 3.

Now change one number and nothing else. With k = 10 the window at minute 9 covers minutes 0 to 9, both earlier accepts are still inside it, the cap of 2 is already met, and the third request is dropped. Answer: 2. That pair of runs is the whole point of the exercise. Two examples differing by one constant prove you actually parsed "window of k consecutive minutes ending at time_i" instead of reading it as "everything so far", which is the misreading this prompt is built to produce.

The routine, in the order you run it

Read the end first

The return sentence and the constraints block carry more meaning per word than anything above them. They give you the output shape and the target complexity. n <= 10^5 rules out the nested loop before you have finished reading the story.

Rename every indexed thing once

requests[i] = [from_i, to_i, time_i] becomes "one request: sender, receiver, minute". Subscripted variables are the specific thing that does not survive a second pass, so convert them to words on the first pass and never decode them again.

Write the three bullets before any code

Given, do, return. If one bullet will not come out, that is the sentence to re-read, and it is usually one sentence rather than the paragraph. Type them as a comment so they stay on screen while you code.

Build your own example, not theirs

Theirs demonstrates the happy path. Yours should target the boundary clause: the empty input, the equality case, the item that falls out of the window. Three elements is usually enough to expose it.

Close the prose and do not reopen it

Once your restatement and your worked example agree, check them against the constraints block only. Re-reading the paragraph mid-solve is where the minutes go, and it rarely returns anything the bullets did not already hold.

In a live round, say the restatement out loud before you code. "Let me play that back: I walk the requests in order and accept one if the receiver is under their cap in the last k minutes." Interviewers score that as strong practice. Nobody hears it as a reading difficulty, and it gets a misread corrected in ten seconds instead of forty minutes.

Eight accommodations for a timed online assessment, and the exact wording

Send these to the recruiting coordinator who owns the invite, not the hiring manager, and send them at scheduling. Column three is generalised from how assessment platforms are built rather than from any measured figure, and platforms differ, so read it as a guide to where pushing is worth it.

The ask	Send exactly this	What usually happens
1. Extra time, as a multiplier	"I read long problem statements more accurately with additional time. Could the assessment be issued with 1.5x time? I understand this normally has to be set before the invite is generated, so I am asking now rather than when it arrives."	The highest value ask on this list. Most platforms expose a per-candidate time setting the employer controls. Naming a multiplier gets a decision, where "some extra time" gets a reply asking how much and loses two days.
2. Confirm what proctoring is on	"Could you tell me which proctoring settings are enabled: webcam, screen recording, tab tracking, or a lockdown browser?"	Not an accommodation, and almost always answered. Ask it first, because the answer decides whether asks 3 and 4 are even possible. A lockdown browser can block other applications no matter what a recruiter has agreed.
3. Read-aloud or text to speech	"I use text to speech for long prompts. Could you confirm that a read-aloud tool is permitted, and that it will not be flagged by the proctoring configuration?"	Two questions on purpose. Permission is the easy half and is usually given. The technical half is the one that fails silently on the day. Get both answers in writing, in the same reply.
4. Restyle the problem text	"Could I have the problem statements as plain text or PDF alongside the platform, so I can set my own font, spacing and background?"	Often declined, because statements are generated inside the platform and releasing them early is treated as a leak. The fallback that usually lands: ask whether the editor pane supports font size, line spacing and theme changes, then set them during the practice question.
5. One time budget, not per-question timers	"Could the assessment run as a single overall time budget rather than a countdown per question?"	Purely a platform configuration question, so the honest answer is that it depends on the tool. Worth asking anyway, because a visible per-question clock costs reading accuracy on exactly the prompts that need a second pass.
6. Spelling not scored	"For any written or free-text sections, could you confirm that spelling and grammar are not part of the score, and that a spell checker is permitted?"	Usually a quick yes for engineering roles, because it was never being scored. The value is not really the permission, it is deleting a background worry during the design write-up section.
7. An untimed practice question	"Is there an untimed practice question on the same platform I can run beforehand?"	Frequently already exists and costs the employer nothing to point you at. Use it to set the editor font and to learn where the run button, the test cases and the timer live, so none of that is new while the real clock runs.
8. Take-home instead of a timed test	"Could I complete the take-home version with a longer calendar window and the same scope, rather than the timed platform test?"	The biggest change and the least likely yes, because many pipelines only have one path. Ask early, before an invite exists. Where a take-home track does exist, moving to it removes the reading pressure rather than softening it.

The email that gets a yes

Subject: Assessment adjustment request, [Role], [date]

Hi [Name],

Thanks for arranging the online assessment. Before the invite is generated I would like to request two adjustments. Is there an accommodations contact I should copy in, or are you the right person?

First, I read long problem statements more accurately with additional time, so could the assessment be issued with 1.5x time? Second, I use a read-aloud tool for long prompts. Could you confirm that it is permitted, and also which proctoring settings are enabled, so that I know whether it will run?

Neither changes the problems I am solving or how the work is judged.

Best regards,
[Your name]

Two asks, mechanics only, no diagnosis and no apology. The last line does the real work: the unstated concern on the other side is fairness to other candidates, and answering it before it is raised stops your email being escalated to someone slower. Then re-confirm the day before the window opens. The commonest failure here is not a refusal, it is a granted setting that never reached the invite.

Where a copilot helps, and where it is the wrong tool

We sell a live assistant, so here is the boundary stated plainly. Software helps with re-reading and with parsing a wall of indexed prose. It does not replace a process change, and there are cases where reaching for it costs you.

It is not a substitute for a granted adjustment. If the real problem is a countdown that is too short, ask for the time multiplier. Sitting the assessment without the adjustment you were entitled to request leaves the next round with the identical problem.
A second block of text can add reading rather than remove it. That is the specific risk for this audience. A generated explanation is more words on a screen that already has too many. Used well it replaces the prompt paragraph. Used badly it sits beside it and now you are reading both.
It cannot help where you have been asked not to use tools. The overlay is excluded from screen capture at the operating system level, so it does not appear in Zoom, Teams, Meet, OBS or screenshots. That is a privacy property, not an anti-detection one. It does nothing against a webcam, a lockdown browser, a proctor in the room, eye tracking, or a phone camera pointed at your desk. Where a process forbids assistance the accommodation request is the route, and the point of this page is that the request is usually granted.
Windows 10 and 11 only. No macOS, no Linux, no mobile, no browser extension. If your interview machine is a MacBook, this is not an option at all.
Not for a live round you intend to talk through. Reading and talking at once is the hardest combination there is, and the restating routine above beats any screen you have to consult mid-sentence.

Where it does earn a place is untimed practice, the part where you build the restating habit across fifty prompts rather than one. Code Copilot in OfferPilot AI captures a problem spanning several scrolls with Ctrl+Alt+S: drag a box, scroll, drag again, then Enter sends the shots together, and it returns the intuition, a dry run and the Big-O rather than only a block of code. Gemini is the recommended model for LeetCode-style problems, and all three models answer in parallel, so switching between them is a display toggle that re-asks nothing. Compare its dry run against your own three bullets. Where they disagree is where you misread, and that is the diagnostic worth having. The free plan is five minutes a day with no card, enough to find out whether a second screen steadies you or just adds words.

Coding interviews with dyslexia: cutting the reading bottleneck