Google IT Support common mistakes
Google IT Support · Understand the traps and how to avoid them.
On a scenario stem like "my laptop is slow," the candidate's first move is to name a specific fix (e.g., "install an SSD") instead of first asking clarifying questions, reproducing the issue, or establishing a theory.
Candidates treat every item as a recall question and want to demonstrate technical knowledge fast, so they jump to remediation even when the question is really testing the order of the troubleshooting methodology.
What to do instead
Before choosing, label which stage the question is asking about: identify the problem, establish a theory, test the theory, plan and execute, verify full functionality, document. Then pick the option that advances exactly that stage and eliminates any option that skips ahead.
When the scenario includes an angry or non-technical user, the candidate picks a technically correct option written in jargon with no acknowledgement of impact, and skips options that include empathy or expectation-setting.
Study time goes into technical content, so soft-skill answer options look like filler and get discounted as "not the real answer."
What to do instead
If the stem contains emotion, a deadline, or business impact, first eliminate any option that blames the user, uses unexplained jargon, or makes a promise that cannot be kept. Prefer the option that acknowledges impact, states what will happen next and when, and explains the resolution in plain language.
The candidate answers "DNS" for every "cannot reach the website" stem, without checking whether the stem shows that pinging by IP works but the name fails, or the reverse.
DNS is the most memorable network service, so it becomes the default explanation for all name-related and connectivity-related symptoms.
What to do instead
Build a one-page symptom-to-service table and rehearse it: no link light means layer 1 cabling or NIC; IP ping works but hostname fails means DNS; local resources reachable but internet not means default gateway or routing; 169.254.x.x address means DHCP failure; one site fails but others work means the remote service, not the local network.
The candidate answers subnet mask, host-count, and "are these two addresses on the same subnet" questions by pattern-matching on the familiar /24 instead of converting the mask to binary.
Subnetting math is the least practiced skill on the exam and is often guessed from memory of the most common home-network configuration.
What to do instead
Practice the arithmetic until it is mechanical: convert the prefix to octets, compute usable hosts as 2^(32-prefix) minus 2, and derive the block size from the last nonzero octet to find the network and broadcast addresses. Write the octets out even when the item looks easy.
The candidate picks "give the user full control" or chmod 777 as the remedy for any access-denied or permission-error scenario.
Full access visibly "solves" the symptom, so it feels like the safe, complete answer, and the security cost of the change is easy to overlook under time pressure.
What to do instead
Answer with the minimum permission that satisfies the task. Memorize r/w/x as 4/2/1 and read octal strings right to left as other, group, owner; know that ownership (chown) and permission bits (chmod) are separate fixes; and remember Windows uses ACLs with inherited permissions rather than a single mode string.
The candidate mixes platform-specific tooling, choosing ipconfig or Task Manager for a macOS/Linux stem, or "/etc/hosts" edits and "sudo" commands for a Windows stem.
Most hands-on practice happens on one platform, so the commands of the familiar OS are the ones that come to mind first.
What to do instead
Keep a two-column mapping table and review it before the exam: ipconfig versus ip addr/ifconfig; Task Manager versus top/ps; Disk Management versus df/fdisk; Device Manager versus lspci/lsusb/dmesg; Event Viewer versus journalctl. Read the stem's OS cue word before looking at the options.
Faced with "a new employee cannot open the shared project folder," the candidate resolves it by adding the account to an administrators group or granting organization-wide access.
Administrator membership appears to guarantee access, and the distinction between identity, authentication, and authorization gets blurred under pressure.
What to do instead
Default to least privilege: identify the specific resource, the specific group that already has the required permission on that resource, and add the user there. Treat any option that escalates the account's global role as a distractor unless the question explicitly asks for administrative capability.
The candidate treats hashing, encryption, and encoding as interchangeable, for example stating that a stored password is decrypted to verify a login, or that hashing protects traffic in transit.
The three terms are used loosely in everyday speech, and exam options are written to sound plausible in either direction.
What to do instead
Anchor on reversibility and purpose: hashing is one-way and gives integrity plus password verification (compare salted hashes; never decrypt); encryption is reversible with a key and gives confidentiality at rest or in transit; encoding such as Base64 is reversible with no key and provides no security. Classify the scenario's goal before choosing.
To make a blocked application work, the candidate selects an option that disables the firewall, turns off antivirus real-time protection, or disables UAC.
Disabling the control does remove the symptom, and answer options are worded to sound like decisive troubleshooting rather than a security regression.
What to do instead
Prefer the scoped remedy: an allow rule or exception for that application, port, or user, or fixing the underlying permission. Treat any option that removes a security control wholesale as wrong on a security item, and as at best temporary on a troubleshooting item.
The candidate speeds up troubleshooting by pasting real logs, internal hostnames, IP addresses, or credentials into a public AI chatbot, or runs AI-generated commands on a production system without checking them first.
The tool is used as a fast search box, and redaction or verification feels like unnecessary friction when a problem is urgent.
What to do instead
Abstract before prompting: replace hostnames, IPs, usernames, and any sensitive data with placeholders, and never paste credentials or customer data into a tool that is not approved for it. Treat generated commands as a draft: confirm each flag and side effect against documentation, and test on a non-production target before applying.