Chances are you’ve stumbled across Google’s new AI search tool in the past few months if you’ve used the search engine recently; it often provides a great overview (except, sometimes, it’s hilariously off-target).
Though the tools evolve daily, provisional as they might be, it looks as though generative AI search engines are here to stay. Since Google’s AI-powered search went live, it’s already made great strides in pertinence and accuracy.
I am so happy to avoid having to scroll through a page of search results to find the same answer – I look forward to see how it will continue to develop.
Still, as a writer, I’m eager to find out what this means for the future of content online. What types of writing will come out of this, and what will we need to do differently as we learn to write around AI-powered search?
Here we address the answers we already know are true about writing for generative AI – and we anticipate other questions that we don’t yet have the answers to.
How AI Search Has Changed Content Marketing
1. New Information Retrieval
Sufficiently large generative engines now so thoroughly circumvent the way users access the information they’re seeking that they should be considered entirely new practices of reading itself.
Rather than lead a user to sources to explore further, generative engines are making items of knowledge available in the form of summaries that can answer questions in a fraction of the time.
I was trying to schedule a movie date for a friend to watch A Quiet Place: Day One. My friend is going on a trip and I wanted to check online if the film would be in viewing theatres when she returned from her vacation.
I searched “how long are movies usually in theaters.”
I got an answer without clicking on a single page.
2. Less Spam
This spring, Google and other search engines began to penalise sites for posting what they characterised as spam or copycat content – a lot of which was AI-generated.
Spammy content isn’t favoured by AI search engines, which are now on the hunt for high authority, well structured articles that the bots can easily scan.
The end result is that what really matters most of all – quality – is still ultimately what really matters the most (even though many others would argue that it should always be this way).
Let’s talk about one of my favorite examples: recipe blogs.
Before, if I searched Google for a recipe for pad thai, an article that consists of the phrase ‘pad thai recipe’ a few dozen times would appear at the top.
I’d wade through an endless list of personalised back of the fridge stuff as if it were a tracking shot across a soccer pitch: here’s a mother in Amsterdam with her new baby, here is a dog, here’s a stock market crash, here’s a marmalade… But no tips on actually cooking the thing.
Now, the top recipe is from Recipe Tin.
Image Source
I still have to scroll before getting to the recipe, but the text is pure usefulness. Nagi talks about why pad thai is or isn’t authentic, the brands she likes for ingredients, and the techniques.
This information helps me cook the dish better.
What I don’t have to do is trawl through spam, you know, because I know it was a human who wrote this for me. I get real tips.
And note: this isn’t to say that you shouldn’t use AI as a supplement to whatever content you produce – only that the content needs to be useful to your readers.
3. Lower Web Traffic
As more and more of these AI-generated answers start appearing in search results, websites have started to report lower levels of traffic in general. For the top-ranked websites, traffic dips of up to 10 per cent have been reported.
This is logical given that searches are dropping through at the summary and not continuing on, but it also means that the goal of your content might not be to rank, but begin to be used by AI and quoted in a summary.
I’m not entirely immune to this even as a creator: I thought about running a half marathon, and wanted to see what sort of training plan I might get.
So I simply googled. What would the experts recommend? No. Instead, I said a bit about my ability and constraints, and ChatGPT made me a plan.
That was a much easier experience that took way less time than looking for an answer.
4. A change in traffic source.
This point above does have a bit of nuance: most content will see some amount of erosion by the date-proving trick, but most of the highly visible content to most LLMs will see traffic on that content increase by 40 per cent or more.
That’s because the aim for an AI searchable text is to be a source, something that people can cite when they write their articles so that, when you’re cited on Google’s AI Search’s summary, or whatever it’s called, they can easily click on your content when people want to read more.
Meanwhile, one of my HubSpot editors, Kaitln Milliken, had her dog spayed shortly before I wrote this, and came right here to the AI-powered search engine, typing in: ‘How do I care for my dog after she is spayed?
‘These tips the AI algorithm gave as an overview were helpful, but I wanted to know more, so I then clicked on the articles that the overview listed first – half of it was because it was easy, but also trusting that this was the authority.’
5. De-emphasis on SEO.
Good for search engines =/= good for AI bots. 2. In many organizations, the incentives are more complex than good journalism is great for traffic. Imagine that you’re a low-level editor at the online news site, say, ‘Meet the Mediocrity’. Your incentives might be designed such that publishing a clickbait piece dedicated to, say, ‘News Anchor Shaves Head to Raise Money for Cancer’, not only won’t harm the site, but it could actually earn you bonus vacation days at the end of the quarter.
Unlike SEO, which is all about keywords, backlinks and rankings, GEO centres on very heavily structured articles full of sourced content. This guide about GEO is here to help you understand the key differences.
How to Write for AI Search
With this roadmap in mind, here’s a few steps toward writing for, rather than against, AI search.
1. Create your page title with priority keywords first
2. Structure your content under subheads created specifically for the search keywords
3. Include as many internal links as you can, especially to demonstrated expertise pages.
4. Limit your writing to around 600 words.
5. Maintain a good ratio of search keywords to natural wording to avoid sounding spammy.
6. Include a search keyword in your last paragraph. Now, you should have a solid foundation for how different search keywords work on your website.
I also reached out to Kaitlin Milliken, who assigns hundreds of articles a year for the HubSpot blog.
Step 1. Find original ideas with high search traffic.
If I am asked to write a blog, and I don’t get a brief, I will start by looking for an original idea around a topic.
For instance, in an ideal world I would go to a medievalist and it would be a low-competition, high-traffic search term, but ultimately I go with: ‘Well, yeah, there’s 20,000 people typing “medievalist”, you’re fine, just get traffic to it, lad.’
If the brief’s topic isn’t original, perhaps I can find a contemporary slant on it. I can insert new information here. I can connect two phenomena that haven’t been tied together before.
As Milliken told me, ‘today it’s all about EEAT, expertise, experience, authority and trust’.
Yes, AI can supply the best practices, but it is from lived, human experience that fabulous content emerges.
‘We know what questions people are Googling so that’s always on our assignment docket, but I want the writer to bring a fresh, trustworthy voice to it.’
She describes a hypothetical blog post about solving customer service.
‘You’re the customer service rep. You’re also the one writing and sending out the articles. What’s your go-to de‑escalation tactic? Okay, now tell me a story about why that works. A bot can’t do that. Bots don’t tell stories. That’s what readers want.
Step 2. Research, research, research.
Then, I begin digging into research. I want to find primary sources with hard statistics and data to give my article some weight.
Sure, you can’t control what you’re writing about sometimes, but make sure that there’s some new information in there to lift your article above others and answer the AI search-bot query thoroughly.
‘But at HubSpot, we see that we do well in search as long as we have original data from our own internal surveys that we conduct. Our writers will take that data, plus any of the newest data out there on the web can be used. So, that gives you trustworthiness; that reads better with trustworthiness and AI search.’
Step 3. Organize your content clearly.
The secret to writing for AI search is a great structure. If I do the research first, the thematic girders are obvious, and I never even need to go back and rewrite.
I might jot H2s and then try to identify places to add lists – often there are pieces of the article that lend themselves to lists, so I will sift through the copy and write a few of those down. A good writer will often want to spread out a page, to give the eye room to breathe.
Behold, for example, the article you’re reading right now. There are a number of subheadings and lists I’ve ordered.
Step 4. Ask experts.
Another way to make sure that you aren’t confused for an AI bot skimming through content that it has already read is to include new or first-hand expertise information. This is a good SEO practice as well, and helps your article to rank higher.
Send some emails to experts asking for their opinion, tip or advice, which your readers will appreciate, and the AI bots will love.
Part of the reason I asked Milliken to write about this for me.
Step 5. Use unique and clear language.
Finally, avoid literary clichés and overused words. Instead, focus on straightforward language.
That is another significant editorial step – remove all the fat and complexity. AI engines like material that is singular and clear. Easier to parse means easier for an engine to produce a reliable summary.
Step 6. Measure your content’s performance.
HubSpot has added an AI Search Grader microapp, which you can use to monitor the performance of content. This microapp is one of the first available tools that supports tracking how content performs on an AI search engine.
7 Helpful Tips on Writing for AI Search
Most of the best practises in AI search are yet to be defined, and it’s going to be a while yet before AI search theory comes of age, but there are already some useful rules of thumb for AI search writing. Here are seven.
1. Provide comprehensive coverage.
Generative engines favour material that gives an overview and answers questions the reader might ask next, so martial a bit of space and cover each topic as fully and from as many angles as possible.
Remember what the AI bot is trying to do – know how to summarise what the user asked to give a useful and relevant answer. The more your stuff is complete, the easier it is for the AI bot to do its work.
2. Cite your sources.
Credibility appears to be the central litmus test for what’s motivating AI bots, and citing your sources is essential.
Use primary sources and statistics where possible, and link to where you got your information from.
3. Lean on industry experts.
Novel, human ‘noise’ information seems to get caught by AI bots quite easily so include quotes and references of expert opinions and advice.
Google is not penalising AI-written content itself, but it is penalising content that is not information-rich. Lean on the expertise of your subject matter experts (SMEs) to ensure your content stands out to your readers and AI bots.
4. Emphasize scannability.
Structure and scannability, it seems, are two of the most critical features of whether or not your article is amenable to AI search. You, therefore, must:
Incorporate lists whenever possible
Limit large chunks of text
Prioritize H2s and H3s that guide the reader through the text
Also, since AI bots appear to have some common favourite text structures, it doesn’t hurt to play around with formats and structure. Again, not confirmed but worth trying.
5. Write conversationally
When constructing the AI’s paraphrase, it helps to have its sources to hand in conversational-sounding form.
Avoid using industry speak where possible, and keep your writing in 7th grade reading level in most cases. Your writing needs to be easy to read for your users as well as AI bots.
Leigh McKenzie, an SEO expert, shares his insights on this tip.
‘To write in B2C language is more valuable now, as we live in a world of increasingly antiseptic, robotic-sounding automated content, and we’re desperate for human voices.’
‘Stories, personal commentary and reporting linked to everyday problems are the sort of content that readers identify with, and that is a trend we will see more of.’
6. Use AI writing tools
Writing tools can help you reach greater visibility through SEO and GEO.
Tracking your metrics can show you whether your content is being used by crawlers, and microapps can help suggestions for search engine optimisation and even grade GEO.
7. Keep up to date on AI search information
AI search best practices are a moving target. While these pointers are evolving as ways to optimise your AI search, be sure to check back for AI search engine updates.
According to McKenzie: While our teams aren’t really robots, staying informed about what’s happening in AI search is more crucial than ever in keeping digital content fresh.
I’ve learned that willingness to adapt is one clue to surviving the ups and downs of the generative SEO scene, says McKenzie: ‘SEO is ever-changing – and you need to adapt.’
Writing for Bots and for Humans
So whenever something shakes up digital content ecosystems, I get like a kid in the Christmas store. Exciting new things to learn and know. That’s why I’ve been paying very close attention to AI – it has the potential to be transformative.
Play with test formats. Throw in some expert quotes. Toss in different kinds of structures. Let your hair down and experiment.
Since AI changes day after day, I’m super excited to see Content be consumed digitally a year and maybe two years from now: Will we still be using SERPs? Will we still search on the net? Will we be only using AI engines?
I’m not sure — but I’m excited to see what happens.
