Using ChatGPT to write gadget reviews without sounding robotic

Every tech reviewer with a browser tab open to an AI assistant has hit the same wall. The first paragraph reads well, the second feels tidy, and by the third the reader can smell the machine. The problem is not the tool. The problem is the prompt.

Most people open ChatGPT, type "write a review of the Sony WF-C700N", and wait. Out comes a tidy, colourless wall of adjectives that could describe any earbud released since 2018. The model defaults to a friendly midpoint because that is what it has been trained to reward. It is the prose equivalent of a sandwich with no filling.

A gadget review earns attention through texture. The slight wobble of a hinge, the way a smartwatch behaves on a wrist during a Brisbane summer, the volume jump between track four and track five on a Sydney commuter train. None of that lives inside a spec sheet.

This piece walks through a workflow that uses ChatGPT as a fast junior writer. You feed it raw material, steer the voice, and rewrite the bits that drift into corporate mush. The end result reads like a person who actually owns the gear.

Open the box before you open the chat window

Every convincing review starts with hands, not prompts. Take the device out, weigh it, plug it into a Type I Australian power board, and pair it. Jot down what surprises you. Does the case feel lighter than the listing suggested? Does the screen wash out under the midday sun you get in Perth? Those notes are the gold a model cannot mine on its own.

Pass them to ChatGPT and you give it anchors that prevent drift. Ask it to describe a fictional pair of earbuds and it reaches for "crisp highs and deep bass". Ask it to describe a pair on your desk after a week, scuffed from a Melbourne tram commute, and the writing has nowhere to hide. A gadget listed at US$199 often lands closer to A$349 here once GST and freight are added, so overwrite the assistant's dollar figure with the number you saw at JB Hi-Fi.

Teach the model how you sound before you ask for anything

Voice is harder to fake than vocabulary. Paste two or three of your own past reviews into the chat and ask the assistant to extract a style profile. Short opener. Long second sentence when explaining a flaw. Specific brand callouts. The model mirrors the rhythm well when given a pattern.

Lock that profile in as a system prompt or paste it at the top of every conversation. Tell the assistant what to avoid as well as what to do. No "in today's fast-paced world". No clunky paired constructions. No three-sentence paragraphs built from the same template. This is where most reviewers quit too early. They assume the tool should know the rules. It will not.

Push for scenes, not adjectives

Generic adjectives are the tell-tale heartbeat of a robotic review. Sleek. Premium. Powerful. Seamless. They mean everything and therefore nothing. Delete every word that could apply to ten different products in the same category.

Replace them with scenes. Instead of "the noise cancellation is impressive", describe a conversation at a busy Adelaide cafe where the barista's steam wand disappears behind the audio. Instead of "the battery comfortably lasts a day", write about forgetting the charger at a mate's place in Hobart and still making it through a Sunday at the cricket. Scenes carry the geography and the social texture that tells a reader you live where they live. A prompt such as "rewrite this paragraph using a moment from a typical Australian weekday" forces the output into detail that survives editing.

Wire in local buying realities

Australian shoppers behave differently from readers in New York or London. Big-box retailers like JB Hi-Fi, Harvey Norman, and Officeworks shape what people actually buy, and their pricing often differs from the sticker on the box. Mention them by name when they are relevant. A line about a A$30 saving on a matching case at Officeworks lands harder than a generic shout-out to "online retailers".

Telco bundles matter too. A smartwatch that pairs cleanly with an Optus eSIM on the NBN talks to a different reader than one that assumes American LTE bands. Warranty is another local quirk. Australian Consumer Law guarantees apply regardless of what the manufacturer writes in the box, and that is worth surfacing whenever a brand pushes third-party repairs. The goal is to write from inside the market, so readers stop translating foreign examples into their own life.

Hunt down the model's favourite phrases

Once you have a draft, read it backwards. That is the fastest way to spot the soft filler ChatGPT loves to insert. Phrases like "it is worth noting that", "that said", and "in summary" almost always signal a passage the assistant wrote without your input. Cut them. Replace them with a verb. A reviewer who says "the hinge loosens after three months" is doing the reader's job. A reviewer who says "it is worth noting that the hinge may loosen over time" is hiding.

Vary paragraph length on purpose. Three short paragraphs in a row feel like a press release. A long paragraph that walks through a single experience, then a short one that delivers a verdict, reads like a person thinking out loud. Aim for that jagged rhythm.

End with a verdict a search engine cannot write

The closing lines are where the robotic voice gives itself away the most. Models default to safe, balanced summaries that offend no one. Readers came for a stance. Give them one. Say who should buy the product. Say who should skip it. Name the competitor you would still recommend instead, and explain why in one sentence only a person who has owned both could write.

If you earn through affiliate links, the disclosure sits here, plain and honest, the way the ACCC expects. A gadget review is, at its core, a piece of advice from one Australian to another. The model can draft the scaffolding. The voice, the verdict, and the lived detail are yours to bring. That is the difference between a review that informs and one that disappears into the feed.