Every Chief Executive, sooner or later, types their own name into Google late at night (and if you are reading this with a faint air of superiority, rest assured that your turn will come, because it comes for everyone), scrolls through the results with one eye half closed, and then clears their browser history as though it might one day be subpoenaed.
Take the Managing Director of a well-established engineering firm, whom we shall call Richard, who has been performing this little ritual every Sunday evening for most of the past decade. Richard knows his first page of results by heart: the LinkedIn profile, the company biography, a less-than-flattering trade magazine photograph from 2014, and, lurking at link nine, a local newspaper report on a planning dispute that he has spent the better part of a year trying to nudge on to page two. He has, in his way, made peace with page one. He knows where everything is and roughly how it got there, and he takes a certain amount of comfort in the knowledge that almost nobody scrolls past link three anyway.
Then, one evening last spring, his teenage daughter suggested he try asking ChatGPT instead, and Richard, against his better judgement, did.
What came back was a single, beautifully composed paragraph that opened with a live planning dispute, credited one of his competitors with the contract he regards as the high point of his career, and closed by describing his firm as “a regional player”, a phrase he has since repeated, with mounting disbelief, to his wife, his Chairman and, on one occasion, a taxi driver.
Well, the planning dispute came from the local paper, which the machine treats as independent and therefore trustworthy. And the contract was attributed to his competitor because the competitor’s announcement had been picked up by a trade title, whilst Richard’s own account sat in a PDF case study that no crawler could read.
These systems give heavy weight to anything they regard as independent, from Wikipedia and Hansard to regulators’ registers and trade press, and slightly less to what an organisation says about itself. They have also, more surprisingly, taken to treating LinkedIn posts by named professionals and threads on Reddit as genuine evidence, so that the question of whether a product is safe or a service worth the money may now be settled, for millions of people at once, by a synthesis in which an anonymous forum contributor and a Tier 1 newspaper sit side by side.
The good news, for Richard and for everyone else, is that most of this can be changed. What a model learns in training is frozen until the next version ships, but its answers also lean heavily on what it fetches from the live web when the question is asked, so a fresh article in the right media outlet or a corrected webpage can start turning up in answers within weeks.
Communications has given the business of responding to all this a name, Generative Engine Optimisation, which is the sort of phrase only a consultant could love, and we say that as consultants.
In practice it comes down to a handful of relatively unglamorous habits. Say plainly on your own website what you do and what you have achieved, in text a machine can easily read, with the answer at the top of the page, using named spokespeople where possible. Keep an eye on the independent sources that describe you and correct them when they are wrong.
Richard, for what it’s worth, has given up googling himself on Sunday evenings. He asks ChatGPT instead, and is pleased to report that his firm is now described as a leading engineering contractor, a promotion that took one rewritten website, two pieces in the trade press, and a polite correction to a Wikipedia entry.
The machines will go on describing all of us whether we take part or not, and it is risky to leave the first draft entirely to them. That is the work we do at Blakeney.
If you’d like to know what the machines are saying about you, get in touch with Kristie Stockdale at [email protected].
Blog