# 53% of AI Users Skipped a Company. Date the Chatbot's Sources First

_Built In found 53% of AI users dropped an employer after asking a chatbot. Here's a 15-minute check: dated sources, WARN, 10-Qs, postings, and team reviews._

- Canonical URL: https://www.refolk.ai/candidates/blog/ai-company-research-check-sources-before-applying
- Published: 2026-11-11
- Reading time: 10 min
- Keywords: using AI to research companies, how to research a company before applying, is this company laying off, research employer before interview, AI job search company research

You asked a chatbot whether a company is worth applying to, and it gave you a confident paragraph. The trouble is that the paragraph is only as current as the newest source behind it, and this month that matters a lot. Canva cut 22% of its staff, Block ran a second round of layoffs, and Fiserv disclosed 3,000 cuts, all in the same week. Here's how to date what the chatbot told you, and how to check it in 15 minutes before you rule a company in or out.

## What Built In's survey says about AI company research

Built In's survey found that AI summaries now decide which employers job seekers pursue and which they drop. Among the job seekers who used AI to research employers, 53% decided not to apply somewhere because of what it told them.

Built In surveyed 119 members of its job seeker community in Q3 2026, and 88% said they had used AI to find or research employers in the past year. The behavior numbers come from the 105 AI users in that group. That's a small sample, so read the results as a direction, not a precise measurement.

| What AI users did after reading AI info about an employer | Share of AI users |
|---|---|
| Decided not to apply somewhere | 53% |
| Responded to a recruiter | 24% |
| Stopped responding to a recruiter or dropped out of interviews | 23% |
| Always or often verify AI employer info | 68% |

> **53%** of AI-using job seekers decided not to apply to an employer based on AI information
>
> Built In, Q3 2026, 105 AI users. Small sample, so treat it as directional.

Built In CEO Maria Christopoulos Katris described it this way: "Candidates are discovering, evaluating and eliminating companies based on information employers may never know they encountered." The word that matters there is *eliminating*. Nobody ever sees the rejection you don't send, so nobody can correct it. When you use AI to research companies, a stale answer costs you most on the companies you quietly drop.

## Why AI company summaries go stale

AI summaries go stale because they lean on third-party sources that lag reality, and because a model can't cite what hasn't been written or indexed yet. A layoff announced this week may not have reached the sources a chatbot pulls from.

Built In's employer report spells out the mechanism. AI prioritizes third-party sources, employee perspectives, and community content over career pages and company blogs. That's reasonable for avoiding corporate spin. It also means the summary rests on reviews, forum threads, and articles that each have their own date, and the chatbot mixes them into one present-tense paragraph.

The errors are real. In a separate candidate study reported by HR Executive:

- 58% of candidates had caught AI giving inaccurate information about an employer, such as wrong salary data or outdated remote work policies.
- Only 16% said the AI response was sufficient on its own.

Employers see the same problem. In Built In's 2026 employer report, 73% of 162 talent acquisition and HR leaders were concerned that AI tools surface outdated or inaccurate information about their company. Only 35% felt prepared to optimize their employer brand content for AI search.

## Nobody at most companies is fixing what AI says about them

At most mid-size employers, nobody's job is correcting a stale AI narrative, so the old version stays up. That leaves the check to you.

An **employer brand role** is a job dedicated to managing how a company looks to candidates: careers content, review responses, and now how AI tools describe the company. In Refolk's index of professional profiles, these roles are rare.

| Segment (current titles, Refolk's index) | Count |
|---|---|
| Employer-brand titles, US | 57 |
| Employer-brand titles, UK | 58 |
| TA leadership titles (Director, Head, VP), US | 2,786 |
| TA leaders per employer-brand role, US | ~49:1 |

These are title matches, not a census, so the ratios matter more than the raw counts. Two things stand out:

- **About 49 US talent-acquisition leaders for every dedicated employer-brand person.** Recruiting has leaders. Reputation mostly doesn't.
- **The UK has about as many employer-brand titles as the US** even though the US workforce is several times larger.

> **~49:1** US talent-acquisition leaders per dedicated employer-brand role
>
> Refolk's index: 2,786 TA directors, heads, and VPs against 57 employer-brand titles.

The US employer-brand titles cluster at large enterprises: Citi, Uber, Siemens, RTX, Honeywell Aerospace, Schneider Electric. Those companies have someone watching what the internet says about them. A mid-size company usually doesn't. So when a chatbot says that company "has a toxic culture" because of an old forum thread, nobody inside has likely pushed back with anything newer.

## The 15-minute check before you rule a company in or out

The check comes down to four questions. How old is each source? Has the company filed or disclosed cuts? Is the team you want still hiring? Do the team-level reviews come from after the latest reorg? Spend roughly the time below on each step.

| Step | Time | Where to look |
|---|---|---|
| 1. Date every cited source | 3 min | Ask the chatbot for links and dates |
| 2. Check layoff filings and disclosures | 4 min | State WARN registries, 10-Q, earnings call |
| 3. Read the last 90 days of postings | 4 min | Company careers page, job boards |
| 4. Match team reviews to reorg dates | 4 min | Review sites, filtered by team and date |

### Step 1: Date every source the chatbot used

Ask for the receipts. Paste this into whichever assistant you used:

> List every source you used for that answer, with the publication date of each. Flag anything older than six months. Then tell me what has changed at this company in the last 30 days.

If it can't give dated sources, treat the summary as a rumor. If the newest source predates a known event like Canva's layoffs, the summary describes a company that no longer exists in that form.

### Step 2: Check WARN filings and 10-Q disclosures

A **WARN notice** is a filing that the federal WARN Act requires before certain mass layoffs and plant closings. Search a state registry, The WARN Database at layoffdata.com (85,000+ US notices across 50 states), or WARNTracker.com (31,977 companies with notices). Search the parent company name too, because filings are sometimes listed under it.

Then check the company's most recent **10-Q**, the quarterly report public companies file. Search it for "restructuring," "severance," and "workforce reduction." Read the latest earnings call transcript as well. A later section explains why WARN by itself won't answer "is this company laying off."

### Step 3: Read the last 90 days of postings

Go to the careers page and sort or filter by date. You want to know:

- Is the team you want still posting roles?
- Are new postings at your level, or only senior and principal?
- Did postings in your function stop on a specific date?

A company that just cut staff and posted your exact role last week is a better bet than one with no layoffs and no postings since summer.

### Step 4: Check team-level reviews against the reorg date

Filter reviews by department and sort by newest. Compare each review's date to the most recent reorg. A glowing review of a team that was merged last month tells you about a manager who may be gone.

Once a company passes, the work moves to the application, and that's where the time goes. Every company you keep needs its own version of your resume. [Refolk](/candidates) does that part for you. Paste the posting, and it rewrites your resume from your own history for that role, drafts the cover letter, and scores how well you actually fit.

## Why WARN filings miss this week's biggest cuts

WARN filings miss many large cuts because the federal law generally covers only employers with 100 or more full-time employees, and only events that hit 50 or more workers at a single site. A WARN search can miss cuts entirely when they're spread over months, spread across locations, or mostly contractors.

This week's cuts show the gap clearly:

- **Fiserv** cut its employee and contractor base by 3,000 since June, including more than 10% of senior management. CEO Takis Georgakopoulos disclosed this on an earnings call. Fiserv hasn't said how the 3,000 splits between employees and contractors. A WARN search won't reliably show five months of cuts with contractors mixed in.
- **Block** cut more than 4,000 jobs earlier this year, then ran a second round without disclosing a headcount. Teams at Cash App and Square were reportedly hit. No number means no clean filing trail.
- **Canva** is cutting 1,271 people, 22% of staff. A US WARN search would only show sites that met the threshold, if any did.

So for public companies, the earnings call and the 10-Q are usually where the real disclosure is. For private companies, the posting history in Step 3 is often the best signal you'll get.

## "Layoffs" is not a verdict on whether to apply

A layoff headline alone shouldn't rule a company out. What matters is which teams were cut and whether your target team is still hiring. Chatbots tend to flatten "had layoffs" into "unstable," and that's how a seeker ends up in the 53% for the wrong reason.

Look at how the market read this week's cuts. Fiserv shares rose about 9.5% in premarket trading after the disclosure. Block's cuts came one day after it reported Q3 results and raised its gross profit forecast. Neither looks like a company in free fall. Both look like they're trimming toward specific priorities, and some teams may keep hiring through it.

> The question is never whether a company had layoffs. It's whether your team survived them and is still posting.

Canva is the sharper example. Melanie Perkins said "our growth has slowed for the first time in a decade," and the company is merging its product, growth, and platform teams. If you're a product designer, that merger could mean fewer, broader roles with more scope. If you were counting on a standalone growth team, that team is gone. A chatbot summary from October can't tell you which of those applies to you. Postings and dated reviews can.

When you find a role that survived a reorg, the posting itself usually tells you what changed, because merged teams tend to write broader requirements. A fit score is useful here. Refolk scores your history against the new posting, so you can see whether you match the post-reorg version of the job or only the old one.

## How to make the chatbot show its work

Don't stop using AI to research companies. Make it tell you its sources and their dates. 68% of AI users already verify often or always, but the HR Executive study found that AI often frames the follow-up searches too, so people end up checking the chatbot against the same sources it used.

Prompts that break that loop:

1. "What is the publication date of your newest source on this company?"
2. "Has this company announced layoffs, a restructuring, or a reorg in the last 90 days? Cite the filing or article."
3. "Which teams were affected, and is there evidence those teams are still hiring?"
4. "Separate what comes from employee reviews from what comes from company filings."
5. "What would change your assessment if it happened last week?"

The last prompt is the most useful. It makes the model say out loud what it can't know, and that tells you exactly what to go check yourself.

## FAQ

### Is using AI to research companies a bad idea?

No. It's a fast way to get a first picture, and 88% of respondents in Built In's survey had used it to find or research employers in the past year. The risk is acting on claims with no date attached. 58% of candidates in a separate study had caught AI giving inaccurate employer information, so treat the summary as a lead to check, not a verdict.

### How can I tell if a company is laying off if there's no WARN notice?

Check the most recent 10-Q for restructuring or severance language, read the latest earnings call transcript, and look for a sudden stop in postings in specific functions. Fiserv's 3,000 cuts came out on an earnings call, and Block gave no headcount at all, so WARN registries alone would have missed much of both.

### Should I still apply to a company that just had layoffs?

Often, yes. Look at which teams were cut and whether your target team posted new roles in the last 90 days. Fiserv's stock rose after its disclosure and Block raised its gross profit forecast the day before its cuts, which points to targeted trimming rather than collapse. When you research an employer before an interview, ask directly how the recent changes affected the team you'd join.

---

_From the Refolk job search blog. More posts: https://www.refolk.ai/candidates/blog/index.md_