Intro to digital marketing - Chapter 14: A.I in digital marketing

Computer vision

An area of computer science that allows computers to “see” an image’s content

Neural networks

Enables a computer to take an image and see if it shows a number

e.g. 0, a 1, a 2, etc

Neural networks (con’t)

White pixels have a value of 0, black pixels have a value of 1, and pixels in between are a shade of gray

The job of neural networks

Find patterns in the 784 values within a 28 × 28 MNIST database that all the 0s have in common

Large language models (LLM)

Software like ChatGPT

Hallucinations 

Incorrect, but plausible-seeming statements based on pattern recognition

Generative A.I

Includes ChatGPT, Google Gemini, MidJourney, and all other Al engines that generate new content.

ChatGPT

Works well when asked for generic ad suggestions

ChatGPT (con’t)

More specific prompts tend to produce better ad copy

ChatGPT (III)

Make sure to request multiple prompts

The three things advertisers need to understand about A.I’s usage in digital marketing

1. AI algorithms need large amounts of data

2. Algorithms can only optimize what they’re told to optimize 

3. AI optimization of ad campaigns can make attribution numbers more misleading

How is it that biased information gets included in AI prompt results

AIs are trained to replicate things seen in documents in documents and other media

Guidelines for using AI in ad creation

1. Prompts should have details about the desired consumer segment being targeted

2. Final approval of responses should be done by one or more humans

3. Generate several ad ideas for each customer segment and ask AI to make highly varied ads

AI and SEO content

AI can plan and add multiple pages of SEO content in a matter of minutes, and can create first drafts of SEO content that can then be edited and revised.

AI and personalization

One-to-one targeting is cheaper, but the marketer won’t gain more information about the customer being marketed to

Guidelines when creating prompts

1. Keep a record of past prompts for future reference

2. Check the Large language model’s temperature of creativity/randomness

3. Specify the response length and thoroughness

Guidelines when creating prompts (con’t)

4. Give examples so the Large language model can best refine the patterns to match

5. Give contextual information to improve the appropriateness of the response to the context it’ll be used in

6. Take baby steps (step-by-step instructions) to improve performance

Semantic search

The intent/reason(s) behind a search, and having A.I determine the intent/reasons

Conversational search

Optimizing voice searches and other search methods with A.I

E-E-A-T (Experience, Expertise, Authoritativeness)

Things high ranking ads must show, regardless if they’re A.I or not

Personalization at scale

Using A.I to create a personalized piece of advertising for a customer