How Generative AI Answers Are Becoming a Discovery Channel for Employer Brands

Most SaaS founders have already typed their product name into ChatGPT to see what comes back. Marketing teams track whether their tool shows up when someone asks for "the best project management software for agencies", and a whole category of tooling has grown up around that question.

Far fewer teams have asked the same assistants a different question: what do you say about us as an employer?

That question matters more than it used to. A senior engineer weighing up an offer, or a product manager deciding which companies to apply to, increasingly starts with an assistant rather than a search results page. They ask something like "which UK fintechs are good places to work for backend engineers?" or "what is it like to work at [company]?" and read a short, confident paragraph. If you are not in that paragraph, or the paragraph is out of date, many candidates will never reach your careers page.

This article sets out why AI answers now sit alongside careers sites and review platforms, what tends to show up and what goes missing, a lightweight audit workflow any team can run in an afternoon, and the fixes that work for SaaS and tech companies without spending on ads.

Why AI answers now sit alongside careers sites and review sites

For years, employer-brand discovery had a fairly predictable shape. A candidate heard about a company, searched its name, landed on the careers page, checked a review site and perhaps looked at LinkedIn. Each of those touchpoints was something the employer could see and, to some extent, influence.

Generative answers add a layer in front of all of them. The assistant reads some of those same sources, summarises them, and hands the candidate a verdict before they have clicked anything. Two things make that layer different from a search results page.

First, it is compressed. A search page shows ten results and lets the candidate choose. An answer typically names a handful of employers and describes each in a sentence or two. Being eleventh on a results page is a weak position; being absent from an answer is no position at all.

Second, it is synthesised. The assistant is not quoting your careers page. It is combining whatever it can find, which might include an old press release, a three-year-old review, a directory listing with your previous office address, or a forum thread. The result reads as authoritative even when the inputs are thin.

Google is explicit that its AI features draw on the same index as normal search. Its documentation for site owners says that to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to appear in Search with a snippet, and that there are no extra technical requirements beyond that (Google Search Central, AI features and your website). Source: https://developers.google.com/search/docs/appearance/ai-features

OpenAI, Anthropic and Perplexity each publish documentation on the crawlers their search features use (OAI-SearchBot, Claude-SearchBot and PerplexityBot respectively), and each notes that blocking those crawlers can reduce how often a site appears. None of the providers explains exactly how an answer is assembled, but together the documentation makes one thing clear: if your employer content cannot be crawled, indexed and understood, it cannot be used.

What typically shows up, and what goes missing

My company, SetpointHQ, publishes a monthly index that measures how AI assistants describe 100 UK employer brands across 14 sectors. For each brand it sends a fixed set of candidate-style questions to four assistants (Claude, GPT, Perplexity Sonar and Gemini), records whether the employer is named and how prominently, and runs structural checks on the careers site and external profiles. The methodology, weightings and full results are public in the SetpointHQ Index at https://pro.setpointhq.com/the-index

To be clear about the disclosure: this is our own research, and the brands in it are large UK employers rather than early-stage startups.

Each brand gets a score from 0 to 100 on five dimensions: citation likelihood, schema completeness, knowledge-graph presence, LLM crawlability and social-signal density (which currently covers Glassdoor only). In the cohort published on 1 October 2026, the median composite score was 55.2. The dimension medians are where it gets interesting:

• Knowledge-graph presence: 89.3
• LLM crawlability: 75.0
• Social signal (Glassdoor): 74.0
• Schema completeness: 35.0, with 19% of brands scoring zero
• Citation likelihood: 26.3

In plain terms, most of these employers are well-known entities and their sites can be crawled. Where they fall down is in the two areas they control most directly: the structured data on their own careers pages, and actually being named when a candidate asks an assistant a question.

Two sector figures are worth flagging for a tech audience. The fintech median composite was 51.2 (15 brands) and the consumer tech median was 48.5 (10 brands), both below the all-sector median. My reading, and it is a judgement rather than a finding, is that tech employers often run careers content through JavaScript-heavy applicant tracking pages and rely on employer reputation built through word of mouth, neither of which gives an assistant much structured, quotable material to work with.

From the Index answers I have reviewed, the things that tend to show up in answers are the obvious ones: company name, sector, headquarters city, and sometimes a headline fact such as size or a well-known product. The things that tend to go missing are the details candidates actually care about: which roles are hired where, how remote and hybrid working really works, how progression and training are structured, and what the benefits are. That is my observation from the work, not a published statistic.

A lightweight audit workflow

You do not need a platform to start. The following workflow takes a few hours the first time and less after that.

1. Write the questions your candidates actually ask. List 15 to 25 questions for your two or three most important talent pools. Mix branded questions ("what is it like to work at [company] as a data engineer?") with unbranded ones ("best B2B SaaS companies to work for in Manchester", "UK fintechs with good engineering culture"). Unbranded questions are where most of the discovery happens, and where absence is most costly.
2. Ask four assistants on the same day. Run each question in ChatGPT, Claude, Perplexity and Gemini. Use a clean session where possible so earlier conversations do not colour the answer.
3. Record four things per answer. Are you named? Where do you appear in the answer? How are you described? Which sources are cited? A simple spreadsheet is enough.
4. Mark every inaccuracy. Old office locations, discontinued graduate schemes, wrong headcount, benefits you no longer offer. For each one, try to trace it back to the page or profile it most likely came from.
5. Check the plumbing. Open your robots.txt, and the robots.txt of your applicant tracking system if your jobs sit on a separate domain, and check that the AI search crawlers named in each provider's documentation are not blocked. View the raw HTML of a job page and confirm the job title, location and description are actually in it, rather than appearing only after JavaScript runs. Run a couple of job pages through Google's Rich Results Test to see whether JobPosting structured data is present and valid.
6. Repeat monthly. Answers vary between runs, so one result tells you little. Trends over two or three months tell you a lot.

Fixes that work for SaaS and tech teams without buying ads

1. Make your careers content crawlable. If your jobs live on a hosted applicant tracking platform, check whether it serves content in plain HTML and whether its robots.txt blocks AI crawlers. If it does not, ask your vendor, or mirror key role and team pages on your own domain.
2. Add JobPosting and Organization structured data. JobPosting markup on every live role, with salary, location and employment type where you can, gives machines an unambiguous version of the facts. Organization markup on your homepage, with sameAs links to your official LinkedIn, Wikidata and other profiles, helps assistants connect the right entity to the right description. Validate after every careers-site or ATS change, because these integrations break quietly.
3. Keep your entity facts consistent everywhere. Use the same legal name, short description, headquarters and links on your site, LinkedIn, Crunchbase, Wikidata and any directories you appear in. In my judgement, inconsistency is one of the most common reasons an assistant gives a vague description or confuses you with a similarly named company. For startups with generic-sounding names, this matters even more.
4. Publish specific, checkable employer facts. Assistants need something concrete to say. A page that explains where you hire, which teams are growing, how your remote policy works, what your interview process looks like, and what your benefits actually are gives them material to quote. Generic values copy does not. This is my judgement, informed by which brands score well on citation in our index.
5. Show up where assistants already look. When Perplexity and ChatGPT cite sources for your sector's questions, note which ones they choose: particular publications, best-employer lists, directories, review platforms. Earning a genuine mention in those places is more likely to feed back into answers than another post on your own blog. Again, that is an observation from our probing rather than a documented rule.
6. Look after your review footprint. Review platforms are one of the few external signals that speak directly to what it is like to work somewhere. A profile with recent, honest reviews and visible employer responses gives an assistant a fairer picture than a stale one.

Treat it like any other visibility channel

Product-led teams are already comfortable with the idea that an AI answer can win or lose them a customer. The same logic applies to hiring, with one difference: the candidate rarely tells you they ruled you out. There is no lost-deal report for the engineer who asked an assistant, did not see your name, and applied somewhere else.

The good news is that the fixes are mostly ones tech teams are well placed to make. Crawlable pages, valid structured data, consistent entity data and specific content are engineering and content problems, not ad budgets. Run the audit, fix the plumbing first, then work on the content and mentions that give assistants something accurate to say.

Author: Andy Pirie, founder of SetpointHQ, a UK recruitment marketing consultancy.