AI in the content process, with a line you can see

AED 6,000 to AED 30,000 per month covers most of what we are asked to build. The number moves on how many pieces and of what type, how much subject-matter interviewing is needed, how many languages, and whether governance policy work is included. Anything quoted before those are known is a guess. We scope first, then price, and the scope document is yours whether or not you proceed. These are market ranges rather than a fixed rate card — the proposal carries the real figure.

Not for the writing. We use it upstream — summarising research, processing interview transcripts, drafting outlines, and generating structured variation like product descriptions at volume. The drafting and editing of anything carrying your name is done by people, and every piece is read once more specifically for the generated-text tells before it ships.

For some content it genuinely is the right call and we will say so — bulk product descriptions from structured attributes, for instance. For anything that has to persuade or has to be right, the cost saving is smaller than it looks once you price the checking properly, and the reputational cost of a confidently invented fact is not small.

Google's stated position is that it rewards helpful content regardless of how it was produced, and penalises content made primarily to manipulate rankings. In practice the problem with most AI content is not the tooling, it is that it synthesises what already exists and therefore has no reason to outrank the sources. That is a content problem rather than a detection problem.

Increasingly, yes, and detection tools are unreliable in both directions so this is about reader perception rather than a test. The tells are consistent: repetitive sentence rhythm, hedging transitions, an introduction restating the title, confident claims with no source. Once a reader notices, it colours everything else on the site.

A first pass is useful and the output is not publishable. The register comes out wrong in a way that native readers notice immediately, and in this market that is a real cost. We have Arabic writers rewrite rather than review.

You own what we deliver under our agreement. The underlying legal position on purely generated material is genuinely unsettled in several jurisdictions and we will not pretend otherwise. For anything where ownership is critical, such as a brand asset you intend to register, we would write it fully by hand and say so.

It depends on the content and the market, and the expectations are moving. Our position is to agree a disclosure line with you up front rather than leaving it ambiguous. For regulated categories, particularly healthcare and financial services here, we would err well towards caution.

Yes, and for many clients this is the more valuable piece of work. Staff are already using these tools, usually without guidance, sometimes with client data. A short usable policy covering what may be used, on what material, and what must be checked is worth more than most content engagements.

Any specific claim — a number, a date, a regulation, a quote — gets verified against a primary source rather than against another article. This is where a large share of the apparent time saving goes, and skipping it is how factual errors propagate through an industry.

Somewhat, and less than the marketing around AI suggests. The realistic gain is in research and processing time rather than in writing time, so a programme might go from four substantial pieces a month to six. If you want forty, that is a different service and we are not the right supplier for it.

Yes, and it is a useful starting point. Send a sample and we will mark up what a reader would notice, what claims are unverified, and where the process would benefit from a human step. It is short work and it usually settles the internal argument.

Treat them carefully in both directions. They produce false positives on human writing that is plain and structured, and false negatives on edited generated text, so a score from one is not evidence of anything much. We have seen entirely hand-written technical documentation flagged as generated. If you are using one to police a supplier, be aware you may be punishing clear writing.

This is one of the clearest yes cases. Structured attributes in, consistent descriptions out, at a volume that would be uneconomic to write by hand. The conditions are that the source data is accurate, the output is spot-checked in batches rather than individually, and the descriptions are not making claims that matter legally. For a catalogue of thousands, it is genuinely transformative.

Partly. It is reasonable for variations on a message you have already decided, and poor at deciding what the message should be. The failure mode on social is that generated captions are relentlessly upbeat and interchangeable, which is precisely what does not get engagement.

A good use case, because you are generating variations to test rather than a single answer. Generate twenty, have a person cut them to five, test those. The generation removes the blank page problem and the human step removes the ones that sound like every other marketing email.

Less than the marketing around AI implies, and the honest breakdown is worth having. Research and processing get materially faster. Writing does not. Checking gets slower, because verifying a fluent draft takes longer than verifying a rough one. Net, on substantial content, the saving is real and modest. On bulk structured content it is large.

We are, for anything we deliver, and that does not change because a tool was involved at some stage. This is worth asking every supplier explicitly, because 'the AI generated it' is not an answer your customers or a regulator will accept.

The same way as any content: does it rank for the query it targeted, does it get read to the end, does it produce enquiries. We do not measure it differently because AI touched part of the process, and we do compare AI-assisted pieces against fully human ones on those measures so the trade is visible rather than assumed.