Chat Support Typing Test
Real live-chat replies and written support responses, with instant WPM and accuracy.
Type what a chat agent actually types — live replies written to the customer while they wait, across order queries, refunds, technical faults and escalations. Built for live chat, email support and non-voice roles.
What you will actually be typing
Taken straight from the test. Every one of these is written to a customer rather than about one — second person, present tense, composed while somebody waits.
Thank you for contacting our online store. I understand you want to return an item that did not fit, and I can absolutely help with that. There is no charge for the return and the process is straightforward. First, let me just check the order. Can you give me the order number, or the email address you used to place it? Thanks, I have found it. I can see the order was delivered on the ninth, so you are well inside the thirty day return window. That closes on the eighth of next month, so …
Thank you for reaching out about the suspicious activity on your card. Protecting your account is our top priority, so I have temporarily frozen the card to prevent any further transactions. I can see two charges that you did not recognize, totaling $212.50, both flagged by our fraud monitoring system this morning. I have opened a dispute for these transactions, and provisional credit will appear in your account within 24 hours. A new card will be dispatched today and should reach your registered address within 3 to 5 business days. In the meantime, any recurring …
I am writing to follow up on your escalated complaint, reference ESC-2024-11907, concerning the repeated failure to resolve your account issue across four separate contacts. I have completed the review and I want to give you the findings in full, including the parts that do not reflect well on us. On the facts, your account of events is accurate in every respect. You contacted us on the second, the ninth, the sixteenth and the twenty third. On the first three occasions you were told the matter would be corrected within forty eight hours. On none …
Who this test is for
Live chat agents, email and ticket support staff, non-voice and back-office agents, social media support teams, and anyone facing a non-voice assessment. It is also the right page if you are moving from voice to chat and want to know whether your typing will hold up once several conversations are running at once.
| Durations | 1, 3, 5 or 10 minutes |
|---|---|
| Target speed | 40 WPM net — but see the concurrency model below |
| Scoring formula | 5-character standard word; one standard word deducted per incorrect word |
| Backspace | Allowed (configurable) — matches real chat work |
| Difficulty levels | Easy (everyday chat) · Medium (refunds, faults, disputes) · Hard (escalations, formal written responses) |
| Passage corpus | 30 passages, 12,071 words, averaging 402 words per passage |
| Register | 86 conversational markers per 1,000 words — the highest of any corpus on this site |
| Cost | Free, no login, results never leave your browser |
Work out the speed your concurrency actually demands
Job postings quote a WPM figure as though chat were a single conversation. It is not. You will be running several at once, and every additional conversation divides the time available for each reply. That relationship is arithmetic, and you can calculate what your target role really needs rather than trusting the number in the advert.
The model, with its assumptions stated plainly
Assume an agent produces about 250 words across a conversation lasting roughly 15 minutes. That is 16.7 words per minute of actual typing for a single conversation. Running N conversations at once multiplies that figure by N, because the same hands are producing all of it.
Both numbers are assumptions rather than measurements, and your own will differ — a transactional queue runs shorter and lighter, a technical queue longer and heavier. Substitute your figures and the arithmetic still works. What does not change is the shape: the speed you need scales directly with how many conversations you hold.
| Concurrent chats | WPM required | What that means |
|---|---|---|
| 1 | 17 WPM | Below the typical advertised threshold — comfortable |
| 2 | 34 WPM | Below the typical advertised threshold — comfortable |
| 3 | 50 WPM | Squarely inside the advertised 35 to 50 range |
| 4 | 67 WPM | Above most advertised thresholds; macros start to matter |
| 5 | 84 WPM | Only sustainable with heavy macro use and short replies |
The interesting result is how closely this lands on reality. The 35 to 50 WPM that employers actually advertise corresponds almost exactly to two or three concurrent conversations, which is what most general support queues run. That is a good sign the model is sane, because it was not fitted to that range — it falls out of the arithmetic.
It also explains the thing agents complain about. Going from three chats to four is not a thirty percent increase in difficulty, it is a jump from about 50 WPM to 67 WPM of sustained output, which is past what most people type comfortably for a whole shift. That gap is precisely why macros and canned responses exist, and why queues that run four or five concurrent conversations tend to be the simple transactional ones where replies are short.
Chat is written to the customer, not about them
This is the difference between chat support and the rest of contact centre work, and it is measurable. Counting conversational markers — you, your, I, me — separates writing addressed to somebody from writing recorded about them.
A voice agent types a case note after the call, in the third person, for a colleague. A chat agent types the conversation itself, in the second person, to the customer, while they watch. The vocabulary is simpler — average word length here is 4.33 characters against 5.66 for general prose — but that does not make it easier. It makes it faster, which is the entire point, because a slow reply is visible.
| Level | What it is | Avg word | Markers / 1,000 | Avg sentence |
|---|---|---|---|---|
| Easy | Everyday chat sessions | 4.03 | 100 | 17 words |
| Medium | Refunds, faults, disputes | 4.4 | 88 | 23.1 words |
| Hard | Escalations and formal responses | 4.6 | 69 | 19.1 words |
The customer can see you thinking
Most chat platforms show a typing indicator, and this changes the job in a way that catches out people moving across from voice. On a call, a pause while you check something is normal and barely noticed. In chat, the indicator stops, the screen goes still, and the silence reads as inattention.
The practical consequence is that chat has two different speed requirements running at once. There is how fast you type, and there is how quickly you produce something. Those are not the same, and the second one is what customers actually judge you on.
Experienced agents solve this with a holding line before any gap: telling the customer they are checking the account and will be a moment. It costs five seconds and removes almost all of the frustration that silence produces. Build the habit early, because it is the single largest difference between a chat that feels attentive and one that does not, and it has nothing to do with how fast you type.
What macros do and do not solve
Every chat platform offers canned responses, and new agents often assume these make typing speed largely irrelevant. The reality is more interesting than that.
Macros cover the predictable parts of a conversation — the greeting, the standard explanation of a returns policy, the closing line, the security preamble. Those parts are genuinely handled, and an agent who has not learned their macro shortcuts is typing several hundred unnecessary words a day.
What macros cannot cover is the part specific to this customer: what you found on their account, why their case is different from the standard one, what you have decided to do about it. That is the part that resolves the contact, and it has to be composed while they wait.
So the net effect is not that typing matters less. It is that macros strip out the easy typing and leave you with a higher proportion of writing that is unscripted and has to be thought through as it is produced. If anything, that makes fluent typing more valuable rather than less, because the keyboard needs to be automatic while your attention is on the decision.
How this differs from the call center test
The two pages are separate and the corpus data shows the split is real — 86 conversational markers per thousand words here against 30 there.
This page covers live conversational typing: replies written to the customer in real time, usually across several concurrent conversations, where speed is visible to the person waiting.
The call center test covers documentation typing: case notes, dispositions and escalation summaries written after the contact ends, in the third person, for colleagues rather than customers.
For a blended role, work through both. The register switch between writing to someone and writing about them is a genuine difficulty and is much of what makes blended work harder than either channel alone.
A four-week plan
Around 20 minutes a day. Week two is deliberately built on short runs rather than long ones, because chat work is bursts rather than sustained output and practising the wrong shape produces the wrong stamina.
| Week | Focus | What to run | Move on when |
|---|---|---|---|
| Week 1 | Baseline conversational speed | Easy passages, 3-minute runs | 95% accuracy at a natural pace |
| Week 2 | Reply-sized bursts | Easy and Medium, 1-minute runs, several per session | Short bursts typed cleanly without corrections |
| Week 3 | Sustained multi-chat pace | Medium passages, 5-minute runs without pausing | Speed steady from first minute to last, no late drop-off |
| Week 4 | Complex written responses | Hard passages, 10-minute runs | 50+ WPM net at 97% accuracy for the full 10 minutes |
If your general speed is below about 30 WPM, work through the free touch-typing course first — concurrency is impossible until typing itself is automatic. For the documentation half of a blended role, use the call center test alongside this one, and for the general technique that applies to any text see how to type faster.
Before a non-voice assessment: a self-check
- You clear the WPM your target concurrency needs, from the table above, not just the figure in the advert.
- Accuracy holds above 97% — in chat the customer reads every error you send.
- You can hold 50 WPM for five minutes, which is roughly three concurrent chats.
- Your speed does not drop across a 10-minute run.
- You type without looking down, since you will be reading the customer message as you go.
- You have a holding line ready for any gap longer than about fifteen seconds.
- You have practised the other assessment components, particularly a concurrency simulation if the process includes one.
Frequently asked questions
What typing speed do chat support jobs require?
Commonly advertised ranges are 35 to 50 WPM, higher than voice roles because in chat the customer is watching you type. What is interesting is that this range is not arbitrary — it corresponds almost exactly to two or three concurrent conversations. On the throughput model set out on this page, two chats need around 34 WPM and three need around 50. Four pushes the requirement to roughly 67, which is why four-chat working is genuinely demanding rather than simply a larger version of three.
How many chats will I handle at once?
Two to four is the usual range, with three common for general support and higher counts appearing in simpler transactional queues. It is rarely your choice — the routing system assigns them, and concurrency often rises automatically at peak. The important thing to understand before accepting a role is that concurrency multiplies your typing requirement almost linearly. Doubling concurrent chats roughly doubles the WPM you need to hold the same response time.
How is this different from the call center typing test?
The register is different and it is measurable. This corpus contains 86 conversational markers per thousand words — words like you, your, I and my — against 30 in our call center corpus. Chat agents write to the customer, live, in the second person. Voice agents write about the customer afterwards, in the third person, as a case note. If your role is blended, work through both pages, because switching between the two registers is its own difficulty.
Do canned responses and macros mean typing speed matters less?
Less than you would think. Macros cover the predictable middle of a conversation — greetings, standard explanations, closing lines — and they genuinely save time. What they cannot cover is the part that is specific to this customer, and that is precisely the part that resolves the contact. In practice, macros reduce the volume of typing while raising the proportion of it that is unscripted and needs composing as you go. Speed still governs how many conversations you can hold.
What is first response time and why does it get measured?
It is the gap between the customer sending a message and receiving a genuine reply. It is measured because it correlates strongly with satisfaction and because it is the part of the experience the customer feels most directly. An automated acknowledgement does not stop the clock in most systems, and customers do not treat it as an answer either. Where voice queues measure how long someone waits before the conversation starts, chat measures the wait inside every turn of it.
Does the customer really see me typing?
Usually, yes. Most chat platforms show a typing indicator, which changes the psychology of the job in a way that surprises people moving from voice. A pause reads as inattention even when you are genuinely researching the answer. Experienced agents handle this by sending a short holding message before going quiet, so the silence is explained. It is worth building that habit early, because it costs a few seconds and prevents most of the frustration that silence otherwise causes.
What kind of text is in this test?
Three graded levels, all written from the agent side of a real conversation. Easy passages are everyday chat sessions — order queries, password resets, delivery chases. Medium passages are fuller responses involving refunds, warranty claims, technical explanations and disputed charges. Hard passages are complex written responses: escalation outcomes, data requests, incident explanations and appeal decisions. The corpus holds 30 passages and 12,071 words.
Is chat support easier than voice work?
Different rather than easier, and the trade is usually misjudged. Chat removes accent pressure, lets you look things up mid-conversation, and gives you a moment to compose an answer. In exchange it demands that you hold several conversations at once, keep them separate in your head, and produce written English that is being read rather than heard. People who find voice work draining often prefer chat. People who type slowly usually find it harder, not easier.
What else is in the assessment besides typing?
For non-voice roles the battery is usually typing speed, written English including grammar and tone, reading comprehension, and increasingly a chat simulation where you handle two or three scripted conversations at once under a timer. The simulation is the part candidates underestimate. It tests concurrency directly, and no amount of single-window typing practice prepares you for the moment a second customer replies while you are mid-sentence to the first.
How do I get better at handling several chats at once?
The counter-intuitive answer is to work on raw speed and accuracy first, because concurrency is limited by how much attention each conversation costs you. When typing is automatic it costs almost none, which leaves attention free for tracking who said what. Beyond that, two habits help most: read the whole incoming message before starting to reply, and never leave a conversation silent without a holding line. Both reduce the rework that makes concurrency feel impossible.
Can I use backspace during this test?
Yes, and it is on by default. That matches chat work, where you correct as you type and the customer sees only the sent message. Running one session a week with backspace off is still useful, since it shows your true first-pass accuracy — but on this page the default is the more realistic setting.
Is this chat support typing test free?
Yes. No login, no signup, no limit on attempts. Scoring runs in your browser and your results are never uploaded to our servers.
Sources
Word density, sentence and conversational-marker figures are computed from our own corpora and can be regenerated with scripts/corpus-stats.mjs. Conversational markers count occurrences of you, your, yours, I, me and my per thousand words, applied identically to every corpus, so comparisons between them are sound even though the absolute figure is a rough proxy for register. The concurrency model above is arithmetic built on two stated assumptions — 250 agent words per conversation and a 15-minute conversation — and is presented so those assumptions can be substituted rather than taken on trust. Those figures are deliberately not taken from our own corpus, whose passages were sized to fill a 10-minute typing test and are therefore much longer than a real agent turn. Typing speed ranges are drawn from commonly advertised roles; no industry body publishes a required figure for chat support work.
- Occupational Outlook Handbook: Customer Service RepresentativesU.S. Bureau of Labor Statistics · link checked 21 Aug 2026
- Employment Projections — 2024–2034 news releaseU.S. Bureau of Labor Statistics · link checked 21 Aug 2026