Four products, one old problem
On 5 August, Symphony Talent launched a service that makes employers visible in ChatGPT, Gemini, Claude and Perplexity. Symphony is one of the biggest names in recruitment marketing, providing career sites and media campaigns for hundreds of employers.
When an agency of that size builds a product around it, the question of whether candidates use AI has been answered.
Michael Drayer, responsible for AdTech and media at Symphony, said one sentence that sticks with you. Candidates form an impression of an employer without ever visiting the careers site.
For someone who has been in recruitment and employer branding for fifteen years, that is a seismic shift. The careers site was the endpoint. All campaigns, all content, all euros led to that one page where the candidate decided.
Now that decision is made earlier. In a model's response, before a single click has been made.
That is what we at AI Rebels have been saying since day one. Good that the market is joining in now.
How a model builds your story
Look how such an answer is formed. Kim asks ChatGPT whether you are a good employer for someone like her. The model does not build its answer from your careers site alone. It draws on your own pages, Glassdoor, Indeed, news articles, and LinkedIn. It compares those sources and turns them into a single story.
If your site and those external sources say the same thing, a clear answer comes out. If they contradict each other, the model remains vague, or it chooses the source it trusts more. That is almost never your own site.
What lies behind those four product names
Looking at Symphony's product, you see four components with four names: an index, an accelerator, an amplifier and continuous monitoring. It sounds like a complete system.
Among those names is work that is bone-dry.
Ensuring that there is consistent, quotable information about you as an employer in places that a language model trusts. Ensuring that this information is server-side rendered, so that a crawler can actually read it and doesn't encounter an empty JavaScript page. Ensuring that the narrative is consistent, whether a model visits your own site or an external source.
No accelerator will fix that for you. It's the fundamentals, done well, everywhere the model looks.
what quotable actually means
Citable means something simple: a fact that a model can pick up and attribute to you. “Salaries between X and Y.” “Four weeks of extra leave upon having a child.” “Office ten minutes from Utrecht Central station.” Concrete, verifiable, the same everywhere. A page full of “we offer a challenging working environment” gives a model nothing to hold onto.
The exact same fact needs to be consistent everywhere. On your site, in your job postings, in the places where others write about you. A single figure that differs in three places, and the model no longer knows who to believe.
Do not build for a single model
And then the part that most employers skip. You cannot build for just one model.
Over eight months, ChatGPT's share of AI referrals dropped from 89 to 63 per cent. Claude climbed to second place. Gemini and Perplexity are taking their share.
Anyone who tailors their entire narrative to ChatGPT will find themselves on the wrong platform next quarter. The answer has to be correct whether Kim asks ChatGPT, Claude or Gemini.
What AI traffic is already yielding
I am not going to tell you that AI traffic is already bigger than Google. It is not. In one dataset, AI sent 6,667 sessions, compared to 113,624 from Google. A fraction.
However, the people coming in via AI stay longer and read more: almost 59 seconds compared to 44 with Google. They arrive with a reason. A model has already suggested you as a good match.
That is the difference between a visitor and a candidate who is already half-convinced.
What is still a niche today will be a channel in six months' time. What is still a channel today will be your first impression in a year's time. And what is still your first impression today will determine who even applies for a job in eighteen months' time.
Start with the question that is already there
For the TA lead whose employer branding budget has shrunk over recent years, there is an uncomfortable truth in this. You cannot buy this channel with a campaign. You earn it with information that is accurate, consistent, and dares to quote a role model.
The launch of Symphony is good news for employers. It is easier to explain internally; there is now a large, established agency by your side saying the same thing.
Procuring a service is one thing. Becoming the answer that a model gives is work that only begins after that.
Start with the boring question that counts. Suppose a candidate asks ChatGPT today whether you are a good employer. An answer comes up. The only question that remains is whether that answer aligns with who you are.
Because that answer is already there. What is missing will be filled in. What is unclear will be simplified. And what is nowhere to be found does not exist.
The only choice you have is whether you write that answer, or whether a model fills it in for you.
My name is Bryan Peereboom, founder of AI Rebels. We make employers visible in language models.