The exam gives you no reading material and no data. Every citation in a Task 2 response is therefore either general knowledge or invention, and examiners read a great many of the second kind. “A recent study found that 73% of employees…” is a fabricated fact; it reads as rigour to the person writing it and as fabrication to the person marking it, and once a reader has seen one, every other claim in the essay is read with suspicion.
What an example can legitimately do is make the abstract visible. A claim about congestion charging becomes checkable when a reader can picture central London; a claim about ageing populations becomes concrete when Japan is named. That is genuinely useful — a reader who has seen the claim in operation understands it better — and it requires no source at all, because you are pointing at something widely known rather than citing a measurement.
The line between the two is the verb. “A case in point is” and “this can be seen in” illustrate. “Studies have proven” and “research shows that 68%” claim a measurement. The first pair costs nothing and the second buys a risk you did not need.
“A case in point is” needs no source. “A 2019 study found” needs one you do not have.
A percentage with no source is fabricated, and hedging it makes it worse rather than safer.
Naming London, Japan or the pandemic is pointing at something known, not citing a measurement.
A paragraph that is mostly example has stopped arguing and started describing.
