ChannelPulse
What changed — AI-assisted rounds and AI-run screens
Two unrelated things now share the name "AI interview". One is a round where the company hands you an AI assistant and grades how you use it; the other is a screen where an AI is the interviewer. This lesson separates them, shows who runs each and how sure we are, and sets up the rest of the course.
What you'll be able to do
- Tell an AI-assisted round from an AI-conducted screen, and know which one you have been invited to
- Say which companies run AI-assisted rounds, and which of those claims are confirmed rather than reported
- Explain why most large companies added one AI round instead of replacing their classic rounds
- Ask a recruiter the four questions that decide how you should prepare
"AI interview" is used for two things that have almost nothing in common, and preparing for the wrong one wastes your week.
| AI-assisted round | AI-conducted screen | |
|---|---|---|
| Who interviews you | A person | An AI agent (voice or video) |
| Where the AI is | In your editor or browser, as a tool you're expected to use | On the other side of the call, asking the questions |
| When it happens | Onsite or late screen, next to classic rounds | Usually the very first step, often within minutes of applying |
| What's graded | How you work with the tool: your plan, your control, your checking | Your answers, transcribed and scored against a rubric for a human to review |
| Roles | Engineers first; now product managers, managers and design rounds | Anyone, but mostly high-volume and first-round hiring |
The first four modules of this course are about the left column. The last module covers the right.
Who runs AI-assisted rounds, and how sure we are
The field moves fast and much of what circulates is rumour, so this table keeps two labels apart. Confirmed means the company says so itself, on its own site or on the record to the press. Reported means two or more independent candidate or prep-site accounts and nothing official.
| Company | What they run | Status |
|---|---|---|
| Meta | An AI-assisted coding round in CoderPad with a built-in assistant. Meta's hiring FAQ says candidates are expected to use it; design interviews now include it too | Confirmed (careers FAQ, Oct 2026) |
| A pilot of AI in coding interviews, plus a "code comprehension" round with Gemini; AI fluency described as a requirement for all software engineering jobs | Confirmed as a pilot (on-record, Oct 2026) | |
| Canva | AI-assisted coding replaced its computer-science fundamentals screen for backend, frontend and ML roles | Confirmed (engineering blog, Jun 2025) |
| Sierra, DigitalOcean | Long builds with any AI tool, then a demo and review; Sierra dropped its classic coding rounds | Confirmed (company blogs, 2026) |
| GitLab | AI welcome in technical assessments if you disclose it and can explain your prompts | Confirmed (jobs page) |
| OpenAI | Varies by round: some formats allow AI, others are designed to test you without it | Confirmed (interview guide) |
| Anthropic | Live interviews are AI-free unless they tell you otherwise | Confirmed (candidate guidance, Jul 2025) |
| LinkedIn, Shopify, Rippling, Stripe, Atlassian | An AI-assisted coding or build round, in several shapes | Reported (multiple candidate accounts each) |
| Amazon | An online assessment variant with a repo bug-fix task and a built-in AI | Reported, not official |
Why it's one extra round, not a new loop
In September 2025 interviewing.io asked 67 interviewers, most of them at large tech companies, whether their company had dropped algorithmic questions. None had. A year later the pattern at big companies is consistent: one AI-assisted round added next to a classic, no-AI round (Meta, LinkedIn, Amazon's assessment). The AI-native startups went further and removed the classic rounds.
The reason matters for preparation. The companies didn't conclude that fundamentals stopped mattering; they concluded that their old questions stopped measuring them. Canva says its fundamentals questions were being solved by AI "in seconds". So the new round keeps the old grading dimensions (problem solving, code quality, verification, communication) and changes the material: a multi-file repo instead of an empty editor, several hundred lines instead of thirty, and problems chosen so the assistant can't solve them in one shot.
The three shapes of an AI-assisted round
Prep sites that interview many candidates describe three shapes, and companies are fairly consistent within each:
- Structured. A browser IDE with a built-in AI chat and a fixed menu of models. You're dropped into a starter repo and work through phases: fix a bug, build a feature, make it scale. Meta, LinkedIn and Amazon's assessment look like this. Lesson 2 walks through it.
- Bring your own environment. Your own IDE and tools (Cursor, Claude Code, Copilot) over screen share. Often an empty repo and a seed problem that the interviewer keeps extending; graded close to low-level design. Shopify, Canva and Rippling are reported in this shape.
- Long build and defend. Two or three hours building something with any AI, then a demo, a code review and a discussion of how it would reach production. Sierra and DigitalOcean.
A smaller fourth shape, AI-assisted code review, gives you a pull request and asks you to use the AI to decide whether to approve it.
AI-run screens are already common
The other meaning is no longer fringe. In Greenhouse's 2026 survey of 2,950 job seekers, 63% had faced an interview run by an AI, and 38% had walked away from a process because of one. Lesson 11 covers how those screens score and how to do well in them.
Four questions to ask your recruiter
Formats change every model release, so ask rather than guess:
- Which rounds allow AI, and which don't? Some interviewers ban it for the first phase.
- Structured or bring-your-own? It decides whether you practise in a browser IDE or your own setup.
- Which tools or models? A weaker in-house model changes how much you can lean on it.
- Is there a practice environment? Meta offers one through the candidate profile; use it.