Employee-generated content is a search asset. Almost nobody uses it that way.
Tribeez Insights
Your staff already know the questions. That is the part you cannot buy.
In 2020 a shop assistant was fired for filming himself at work. This year Gap and Starbucks started paying their staff to do exactly that. The reversal has a name, and almost nobody has worked out what it is actually good for.
Tony Piloseno was a college student working at Sherwin-Williams when his TikToks of himself mixing paint went viral. The company fired him for gross misconduct. That was five years ago.
This month Gap opened its creator programme to staff across Old Navy, Gap, Athleta and Banana Republic, letting them earn commission on what they post. Starbucks became the first brand to pilot a custom Creator Network with TikTok, paying employees a share of ad revenue when their videos get used in advertising.
The industry has settled on a name for this: employee-generated content, or EGC. Staff posting about the place they work, on their own accounts, in their own voice. In South Africa it is barely discussed. In the US trade press it is the most asked-about topic of the year, and Keith Bendes at Linqia says he has never had so many reporters call him about a single subject.
His piece on it is worth reading in full, because it is the most honest thing written on the subject so far. He breaks EGC into three models and, crucially, refuses to pretend any of them is free.
Affiliate commission, which is what Gap has done. The existing affiliate programme, opened to staff across offices, stores and distribution centres. No deliverables, no obligation, commission and product when someone buys through their link. The brand pays only on conversion.
Revenue share on ads, which is Starbucks. Announced at Cannes in June, building on the Green Apron Creators programme it launched in 2024. Employees opt in, Starbucks posts briefs to the network, and employees earn a share of ad revenue when their content gets used in paid media. Free to the brand until paid budget is allocated.
Paid content, which is a normal creator relationship that happens to be with an employee. The only one of the three that guarantees anything gets made, and by some distance the hardest to run.
Bendes then lists the questions that make the third model difficult, and they are not marketing questions at all. Are the employees hourly or salaried, and are they filming on or off the clock? What happens when a customer's face ends up in shot without consent? What if the store is shown in a way that breaches policy? What if the filming gets in the way of the job? Does their manager even know what they have signed up for?
His conclusion is blunt. EGC looks cheap and usually is not, and the proportion of employees who produce anything usable is far smaller than brands assume.
He is right on all of it. And that whole conversation, the one South Africa has not started yet, is stuck on the wrong question.
Cost and control are the second question
The first question is what employee content is uniquely good at, and the answer is not "cheap volume". Cheap volume is what brands reach for it hoping to get, which is exactly why Bendes has to keep warning people that the volume will disappoint them.
Here is what your staff have that no agency, no creator roster and no content budget can produce.
They know the questions.
The person on the till, on the phone, on the shop floor gets asked the real buyer questions all day long, in the exact words buyers use to ask them. Does it shrink in the wash. Is the big one actually worth it. What is the difference between these two. Which one do you use. Is this a good price or should I wait.
That is not a content asset. It is a research asset that happens to arrive as content.
Most brands are paying somebody to work out what their buyers ask before they buy. Their own staff could write the list from memory in ten minutes, and would be more accurate, because they are not inferring it from keyword volumes. They heard it on Tuesday.
Why that matters more than it used to
It used to matter mildly, in a customer service sort of way. It matters commercially now, because the questions are where buying happens.
When somebody asks an AI assistant which product to buy, the model does not have an opinion. It borrows one from wherever a human already published theirs. Tinuiti's Q1 2026 AI Citation Trends Report found roughly 82% of AI citations pointed at earned media rather than a brand's own site. Your website is not the input. Other people talking about you is the input.
And the surface set is wider than search. On a South African skincare brand we mapped recently, Google was tracking 12,100 monthly searches for the category. Across all five surfaces where buyers actually ask, the real figure was 35,900, and 66% of that demand had no creator content standing in front of it at all. Google alone under-counted the market by three times.
Sixty-six per cent of demand with nothing in front of it is not a content problem you can spend your way out of. It is a coverage problem, and coverage is where employee content is structurally advantaged.
The long tail is the whole game, and only employees can reach it. A brand can fund perhaps twenty creator videos a quarter. Two hundred employees each answering the questions they personally get asked produces hundreds of specific answers to specific questions. Search rewards the tail. The tail is uneconomical to commission. It is nearly free to encourage.
Employees speak in customer language. Marketing says "colour-lock technology". The person on the floor says "it does not go grey in the wash", because that is what the customer asked. Retrieval matches the second one. Literal language beats clever language, and employees are literal by default.
It compounds instead of spiking. A campaign runs and stops. Staff post continuously, and every question that gets answered stays answered.
The bonus: your media budget stops guessing
There is a second return here that most write-ups miss entirely, and for a performance marketer it may be the more persuasive one.
Think about how paid creative normally works. You brief it, you produce it, you pay to distribute it, and only then do you find out whether it lands. The money is spent before you learn anything. Creative testing is a line item, and a large one.
Employee content inverts that sequence. The content gets made for nothing and posted organically. The feed then does your creative testing for you, at no cost and with a sample size no test budget could buy. You watch which posts actually perform, and you put money behind those. You are not buying creative any more. You are buying distribution on creative that has already proved itself.
This is not a theory anyone has to be talked into. TikTok has built the product. Marketing Dive describes Creator Networks as a way for brands to assemble a customisable pool of employees and advocates, with the explicit aim of using that content in paid advertising. Starbucks is the pilot, and the compensation model is a share of the ad revenue, which tells you exactly what the content is for. TikTok's own figures put Spark Ads at 134% higher completion and 157% higher six-second view-through than standard in-feed ads, and the boosted post keeps the likes, comments and shares it earned organically.
So employee content pays back three times, not once. Organic reach. Coverage of the questions. And a supply of paid creative that arrived pre-tested.
Most brands are counting only the first one, which is why the business case keeps looking thin.
The parts I do not think anyone has tested
Here is where I want to be careful, because there are two real tensions in this argument and it would be easy to skate over both.
Employee content is not earned media. An employee is affiliated. If 82% of AI citations point to earned sources, the obvious question is whether a machine treats a barista's TikTok as a person talking or as the brand talking.
I do not know the answer. I have not seen anyone test it, and I would be suspicious of anyone who claims certainty either way. What I would say is that the answer probably depends on which of Bendes' three models you picked, which makes his framework more important than he lets on. Affiliate and revenue-share programmes leave the content on the employee's own account, in their own voice, indexed as an individual. Paid-content programmes, with briefs and deliverables and required disclosure, look progressively more like brand publishing. The same content can read as a person or as an advert depending on the plumbing behind it. Disclosure rules push one way and retrievability pushes the other, and brands are going to have to hold both.
And the media asset sits on somebody else's shelf. This one is more prosaic and I have not seen it raised anywhere. When you boost an employee's post, you are running paid media off a post on their personal account. If they resign and delete it, or simply take the video down because they have moved on, the ad stops. Your best-performing creative is hosted on infrastructure you do not own and cannot control, held by somebody who no longer works for you.
That is not an argument against doing it. It is an argument for sorting out content licensing and whitelisting at the point where somebody opts in, rather than discovering the problem when a flight is live. Sixty-one per cent of consumers already think brands should compensate employees who promote them on social. Paying properly and papering it properly are the same conversation.
So measure it properly
If Bendes is right that only a small share of employees will produce anything usable, and he is, then counting posts is the wrong measure. It guarantees the programme looks like a failure.
Count questions answered instead.
One employee who answers the fifteen questions your category actually gets asked is worth more than fifty who post a nice shot of the storefront. The commercial unit is a question answered, not a post delivered. That reframes the entire business case, because you are no longer asking employees to be creators. You are asking them to do on camera, once, the thing they already do out loud, fifty times a week.
Where I would start
Ask ten of your customer-facing staff to write down the ten questions they are asked most. Do not brief it, do not shape it, just collect it. That is your question map, and it cost you one team meeting.
Then ask each of those questions out loud, into Google, into ChatGPT, into TikTok, into YouTube, and write down who comes back. For most brands it will be a competitor, a review site, or a creator with no relationship to anyone in the category.
Then decide which of the three models you can actually operate, with your legal team in the room rather than after the fact, and settle the licensing before anybody posts anything.
The brands that get this right will not be the ones with the biggest employee programmes. They will be the ones who worked out that the people answering customer questions all day were sitting on the most valuable, least examined asset in the business, and finally asked them to say it where the question gets asked.
Sources: Keith Bendes, "Gap and Starbucks bring EGC into the spotlight", The Art of Influence, 21 August 2026, for the three EGC models and the operational considerations. Marketing Dive, 23 June 2026, for the Starbucks and TikTok Creator Network pilot; the 3x employee posting rate cited there is brand-supplied. Sprout Social State of Social Media for the employee-content discovery and compensation figures. Tinuiti Q1 2026 AI Citation Trends Report, via Search Engine Land, for the 82% earned-media citation figure; note this is vendor research, disclosed by its author. Spark Ads performance figures are TikTok's own. Tribeez Answerz Gap, South African skincare category, for the surface and coverage figures.
Get The Performance Creator
A weekly read on creators as a performance engine. Social search, visual commerce, and what the platforms just shipped. Free, and you can leave any time.
Subscribe →© 2026 Tribeez Africa. All rights reserved.
