What Should an AI Search Agency Look Like?

A Brief History of Google Updates

Google Florida was the first big Google update, and in the SEO community it got its name because it was rolled out just before Pubcon Florida that year and became the main topic of conversation at the event.

More importantly, Florida was Google’s first real attempt to combat search engine manipulation; it ignored links to your site from irrelevant pages. This changed how linkbuilding had to be done and eliminated one of the early, straightforward ways to cheat your way to the top of the rankings.

Two years later came the Big Daddy update, which adopted a nickname from one of the Pubcon attendees. Google slowly replaced all its old data centres with new ones which processed information differently.

Big Daddy wasn’t designed to combat spam but simply to better push good websites – websites that actually answer the user’s questions and give them what they need to know. Google was looking to whittle down the number of pages it needed to care about, eliminating low quality pages and sites designed to manipulate rankings.

However, this didn’t prevent people from carrying out ‘shortcuts’, tricks which would artificially boost a website’s rankings without making it more valuable to the end user. There are still a few agencies like that in our industry, but the balance tipped in 2011 when Google Panda was released.

Google Panda dealt with webspam and other dubious tactics, and one aspect of how it worked was to introduce penalties. Before, updates had just rewarded sites that gave users a better experience. Panda allowed Google to actively suppress sites from results if they were obviously operating in bad faith.

Panda was later joined by Penguin, SpamBrain, and the Helpful Content Updates, among many others, all of which significantly reduced the web presence of many websites which were using shortcuts to get to or stay at the top.

And that brings us back to AI Search.

What Works to Improve Your Performance in AI Search?

This blog is being written at the tail end of September 2025. It seems likely that at least some of what we discuss in this section will be outdated within twelve months, and possibly even by the end of the year.

You may well have seen people on LinkedIn promising a secret shortcut to the top of the AI search results, and you’ll have seen others say that there isn’t any such thing.

Actually, there have been some very successful experiments. At the moment, posting links to key pages in the middle of essays on LinkedIn, Reddit, or other UGC (User Generated Content) sites can be effective in some fields – but we don’t recommend doing this heavily.

Why not? Because AI search is in its infancy. When search engines were in their infancy, shortcuts and cheap tricks still worked and weren’t penalised.

It’s hard to imagine that the engineers of AI search won’t learn from history. The equivalents of Google Panda – where overusing a technique that doesn’t add value get the site penalised – will not be a decade away. They might be less than a year away.

Alternatively, analysis of pages that are successful in AI Search suggests that most of the websites that are referenced most often also dominate their rankings in traditional search.

This means that, at least for the time being, one of the best ways to perform well in AI Search is to focus on what works for SEO – or rather, for a part of it. It’s been shown that Schema structured data, for example, is not currently used by AI systems. (We still recommend Schema be maintained on your site, however, for the 90% of searches that don’t go through an AI – and the same is true for other technical issues that could affect regular search. It’s not over yet!)

Agentic Behaviour and the Future of AI Search

Before you dismiss the above as not having faith in AI, we do recognise that the field is growing, and that means AI search can have value to our clients.

As such, we keep a close eye on developments – we’re just not going to use techniques for our clients that will see them penalised in the future. It might be easy to pump our numbers that way, but planting a ticking timebomb on a client’s website isn’t how First Digital Media operates.

As such, we keep a close eye on the state of AI search, new experiments, and also on plans the big names have announced. Something to watch for in the near future will be trials of agentic search.

What is agentic search? It’s asking AI agents to handle a multi-step process for you. You might say “Give me four different methods to get a poorly, crying child to sleep, and compare their pros and cons”. In that situation, you probably don’t have time to do that kind of deep research, but the results could make your life much easier.

Once agentic search is performing properly, the likelihood is that we’ll also see agentic commerce.

It might be as simple as “buy me tickets to a Chappell Roan gig in Manchester, don’t spend more than £150”, or as complicated as “Order all the ingredients I’ll need for a red velvet cake, make sure it’s all next day delivery”. The AI will then run through all the steps needed to make the order.

Our prediction is that running paid ads that target agentic commerce will be an extremely effective (if highly competitive) option. It’s easy to see Google, Microsoft, and Meta Ads agencies getting very involved in this. Our eCommerce team are watching developments carefully; whatever reservations you may have, the odds are good that it’s something we’ll have to contend with.

So while we predict that the renamings won’t stick, and that the gurus offering shortcuts will be storing up future damage for their clients, we keep our eyes open. If you’d like a sensible approach to search (AI and otherwise), why not get in touch?