A Black Friday page used to have one audience: a person with a phone and a budget. This year it has a second one. Some shoppers now ask an assistant to find, compare and sometimes buy on their behalf, and that assistant reads your page differently from a human. Many cannot, or may not, squint at a badly formatted price, wait for a script to finish, or work out what “free delivery on selected items” means. If a page is ambiguous to a machine, an assistant may skip it.
None of this requires new technology on your side. It requires finishing the basics that most sale pages leave half done.
What an agent needs from a sale page
When we look at a sale page through the eyes of an assistant, these are the seven things we would check:
- A stable URL. One address for the offer that does not change per campaign, per day or per tracking parameter.
- A price in plain text. The current price, and the previous price if you show one, in the HTML the server sends, not assembled by a script after load.
- Stated availability. In stock, low stock or sold out, written on the page, not implied by whether a button is greyed out.
- A delivery timeframe. Stated per product or per region, not only in a footer policy.
- Returns terms. One clear statement, linked from the offer.
- Structured data. Product and Offer markup that repeats exactly what the visible page says. Our how-to on adding Product and Offer schema to sale pages covers the properties and the checks.
- A checkout that lets a customer’s agent through. A human-verification step that fails when an agent acts for a genuine customer turns a purchase into an abandoned basket, with nothing in your reports to say why. The mechanics and the competing protocols are in our earlier piece on websites blocking AI shopping agents, and we will not repeat them here.
The South African context
The calendar is already moving. Black Friday falls on Friday 27 November 2026, and Takealot said its Black Friday-priced deals run from 26 October to 23 November, with new categories each week (BusinessTech, 23 October 2026). Takealot has stretched the event into a month.
Shoppers started early too. Briefly News, citing Google Trends, reported South African search interest in Black Friday up 30% over the past three months compared with the same period last year. That is a single Google Trends comparison reported second-hand, so read it as direction rather than precision.
On the assistant side, In April, IOL reported Sixty60’s launch of Pixie, headlined as South Africa’s first AI shopping assistant, which recommends products from a customer’s buying history. A retailer-owned assistant is a different thing from a general-purpose agent, but it points the same way: shoppers are getting used to software that does the comparing.
A September Bizcommunity piece on the South African retail peak, written by Incubeta’s Zaida Kamish and Malcolm Peters, makes recommendations that line up with the list above. It urges retailers to keep permanent deal pages live year-round, to mark offers up with complete Product and Offer schema, and to state delivery timeframes. It also says retailers on five-to-seven-day delivery windows risk ruling themselves out unless they offset it with much better pricing. Our how-to on building a permanent Black Friday deals page turns the first of those into steps.
What we do not know
We will be plain about the limits. We are not aware of any published measurement showing that Product and Offer schema changes how an AI shopping agent chooses between retailers. Google documents the markup for its own shopping features, and that is the evidence base for doing it. Whether any given assistant uses it, and how heavily, is not public.
We also cannot tell you how many South African shoppers are using an agent to buy this Black Friday. The adoption data we trust does not exist yet for this market. Anyone quoting a figure is extrapolating.
What we can say is narrower and still useful: ambiguity costs you with machines for the same reason it costs you with people, and removing it is cheap.
What to do this week
Open your best-selling sale page and answer five questions, in this order:
- Are the price, stock status and delivery information in the HTML source, not only on screen?
- Is availability stated in words?
- Does the page state a delivery timeframe for a Johannesburg address and for the rest of the country?
- Does the markup say the same thing as the page?
- Can a customer’s agent get through checkout? Look for a human-verification step, pop-up or interstitial that only a person could solve or dismiss.
If the answer to any of these is no, that is your work for the week. The fixes are mostly content and template changes, not projects.
Then keep the page. A deals URL that exists all year, is updated for each event and is linked from your main navigation keeps whatever links and history it earns, which a campaign page built in November and deleted in December cannot.
If you would like a second pair of eyes on a sale page before the peak, get in touch.