1: What Goes In, Defines The Quality of What Comes Out.

The prompt you pop into your chosen AI system provides the model with guidance on WHAT you need, HOW to aggregate the information and WHICH information to display as an output. If you take the time to carefully steer your AI ship, the results will be more rewarding.

Ensure that your prompts are simple, clear and contextual. If there are particular constraints you need applied, include them - for instance, if you are a teacher preparing class materials, and you need the model to "use language appropriate for an 8 year old", tell it that. 

If there is a particular format you need the AI response to take, be sure to explain clearly. For instance, if you need your AI bud to scan the first 5 chapters of Charlie and The Chocolate Factory for references to "cabbage soup" and list these in a table, tell it how many columns the table should be, and what information to include in, such as the page number where the reference can be found, and a summary of the paragraph where the reference appears.

If there is a common misconception or an easy mistake that the AI could be prone to making, point this out and tell the system to watch out for this. For example, if you are asking the model "when did WWII end?" and you know that there are numerous different dates being bandied about, especially by AI generated responses that are ranking well, or less reputable websites, you can tell AI to use particular references to use the information you need, or false information to avoid factoring in. For example, a good prompt in this mode could be, "using Wikipedia and the National WWII Museum as reference points, tell me when WWII officially ended and what led to the end of the war. Summarise this in bullet points."

2: Burnout Is A Real Thing - Don't Overload Your AI Buddy With Complex Constraints

Don't overload your AI friend with too many constraints and requirements; there is a real risk of confusing the model if you use too many qualifiers, constraints and structures in your prompt, which could break it or leave you with unsatisfactory results.

Think about how you'd brief a member of staff if you needed them to do something; that's the kind of briefing your AI model needs - clear, precise, helpful. Endless sentences containing complex requirements, and contradictory prompt requirements will cause the model to fail and leave you frustrated. You can always use follow up prompts for more information or an additional element to the original response.

3: You Might Have To Spoon feed Your Chosen AI Important Info.

Depending on the model you've chosen and whether you're using the free or paid version, it may not be able to surf or "see" live website content, so if there is a specific site you want crawled and considered as part of the output, you may need to screenshot, copy and paste or summarise the information for your AI friend so that it can do its job. Claude, for instance, which is a popular AI option at the mo, does not have the capability to visit websites on its own if you opt to use the standard version. 

While inputting data to your AI might feel a bit like feeding those 90s money banks pennies - the ones with the cute robot faces and whirring insides - actually, giving it a PDF of your CV to chew on, or a word doc version of your novel first draft to scan for grammatical issues, can be really helpful and well worth the few extra seconds to find and upload the file.

4: AI Can Not Think Like A Content Strategist

Following on from this, it's really important to understand that, while incredibly useful, AI is not and will never be a content strategist. It is an information aggregator that takes requests and churns out responses based on machine learning that is entirely reliant on human guidance and gatekeeping.

• AI can not tell if information is factually correct
• AI can not correctly identify human sentiment and empathise with an audience
• AI can not think faster than the humans who are feeding it information - it can not create, only regurgitate, and it can only do that as well as it has been taught to.

While it's understandable to be nervous of AI and its future role in the copywriting and content strategy spheres, the absolute best thing you can do is to jump in and use the AI systems and tools available to you, so that you can get a handle on the benefits and limitations on this robotic information aggregator. It is true that AI has parsed and regurgitated copy for a wide variety of channels and sectors by now, but it is humans who must prompt, review and gatekeep this copy to prevent misinformation from spreading. My worst fear as a content strategist and web citizen is that we stop being able to tell fact from fiction, that tech companies and businesses stop investing in AI management and monitoring, and that we end up with an internet full of generative AI content that is utterly incorrect, and a population unable to tell that this is the case. Super scary stuff. So, keep creating, keep gatekeeping, keep monitoring and reviewing those systems, because AI does not have the imagination or mentality to gatekeep itself.

5: Generative Search Is Not Going To Replace Organic Search

What generative search does, is collate all the best ranked responses to a particular query, assimilate the information from these web pages, and display them conveniently on the results page, preventing the need for a searcher to click away.

The high ranking content that generative search seeks and scans, is not produced by AI, and is essential to the generative search system being able to supply answers within its algorithm. If Google starts ranking AI-generated content higher than content created by humans, then this would be a concern. However, reputation is still king in the age of generative search, and as long as you focus on securing strong backlinks, gaining social equity and ensuring the reputation of your website stays strong, there is no reason why you shouldn't appear in both generative and organic search results, thus increasing your overall visibility.

The old tennets of writing high quality content that directly answer questions and provide clear solutions are still valid - differences may come in ranking factors like quantity of content - as generative search tends to favour clear, succinct content with direct answers and references over War and Peace style articles. 

6: Training AI To Be Better Is The Right Thing To Do 

Whether you love or hate AI, it's here to stay. The more of us with experience, education and authority who are carefully inputting clear prompts and vetting outputs, the better. Let us leave no room for misinformation, poor quality data and a robotic internet full of generated things. If we feed it well, it will grow up healthy and strong as a helpful asset for content creators, business owners and consumers alike.