AI traffic is paying off on Shopify, but ChatGPT ad visits earn less than paid search

If you build it, they will come. Shopify built commerce into AI agents, and the shoppers came.

Pay ChatGPT to bring them, though, and most just come to watch.

Shopify had a quarter: Q3 revenue grew 32% on $119 billion in GMV.

AI shoppers came to buy: Shopify’s investor slides show AI traffic to stores up 2.6x year over year, and orders from AI searches up 3.3x.

AI-referred sessions converted 64% better than organic search.

Organic search got some breathing room: On desktop, Google moved AI Overviews, its AI-written answers, to the right-hand column for many brand searches.

Paying for AI clicks is another story: A working paper by Maximilian Kaiser and Christian Schulze compared paid ChatGPT and Google search visits on 102 advertiser sites. In US traffic:

• A ChatGPT ad visit brought in $0.89, against $2.45 for a search ad visit.
• Of the 87 advertisers buying both, 73 got zero sales from ChatGPT ads.
• To match search’s revenue per ad dollar, a ChatGPT click would need to cost about $0.26. The largest publicly known account paid $1.72.

The data stops in August, so it’s an early read.

How we’d play BFCM: Think twice before moving search budget into ChatGPT ads. If you test anyway:

• Set your max CPC from your own revenue per visit, not OpenAI’s suggested $3 to $5 starting bids.
• OpenAI’s creative enhancements rewrite your ad for each conversation, sight unseen. They’ve been on by default since Oct 30, even in existing self-serve campaigns. Opt out for a clean read.

Either way, put your AI effort into product feeds, reviews and pages AI agents can cite.

Reddit and TikTok expand search ads as Instagram and X pay creators for video

Social media pulled a Freaky Friday this week.

Reddit and TikTok woke up as Google. Instagram woke up as YouTube, and X as TikTok.

Ask Reddit, get an ad: Reddit is testing ads in its search results with select advertisers.

Sell espresso machines? Your ad could show up for “espresso station” searches, since Reddit picks ads partly by relevance.

Reddit needs the money: Google’s AI search sends it less traffic.

TikTok gives search a FOMO meter: Impression Share and Top 1 Impression Share show what percentage of eligible search impressions your ads captured, overall and in first place.

Both go down to keyword level. Low share means missed demand:

• Spending your full budget? Raise it.
• Money left over? Check your bids.

TikTok’s Keyword Suggestion Tool also gained Keyword Packs: ready-made brand and category keyword lists that refresh daily. TikTok hasn’t named markets for either.

Instagram is still trying to make long-form happen: It’s paying select creators to post to Features, its test of videos up to 90 minutes.

Business Insider reports offers from $50,000 to $500,000, and Meta says it’s funding some creators directly.

Meta also paid creators on IGTV, its old long-form app, and began shutting it down in early 2022.

X hands out attendance awards: Its invite-only Video Kickstart Bonus pays creators who post 15 vertical videos on 15 different days in about a month.

Videos must be their own and under three minutes, but clips they already posted on TikTok, Instagram or YouTube count.

Some of your creator partners could earn twice from one video, so check what your contracts allow on X.

AI Mode searches run 3x longer. Here’s how your ads can keep up

Your customers used to search for “running shoes.”

Now they describe their entire knee injury history.

Ginny Marvin, Google’s Ads Product Liaison, asked Shivani Mohan, who leads Google’s search behavior research, what that means for advertisers.

First, caveman queries are on their way out. In AI Mode, Google’s chat-style search, queries run three times longer than traditional searches, and one in six uses voice, images or video.

Many are questions searchers never thought to ask before.

The Crew’s take: The long tail converts, too. Google Ads data we covered in September showed one- and two-word queries falling from 42% to 24% of impressions since January 2025.

Three- and four-word queries jumped from 20% to 46% of conversions.

Asking twice is a compliment: More searches don’t mean the first answer flopped. Google counts it as a failure when you add words to fix a bad result.

The extra searches are new questions, often follow-ups: check San Francisco’s weather, then ask, “What about Austin?”

Remember, Google now shows ads on AI Mode follow-ups in the US.

Search, Shopping and PMax campaigns qualify by default, with no separate opt-out, so review your negative keywords.

Brainstorming moved into the search bar. Shoppers now ask how to think about a whole category long before picking a product.

Ginny’s advice: judge search beyond the last click, and test more creative assets for that early research.

The summary is the trailer: People still want the movie. For how-tos, that means videos. Before choosing between two products, it means reviews from someone in their exact situation.

So know what customers say about your business online.

Make your ad the obvious next step: Ads in Google’s AI answers match both the query and the response, so start where that response ends.

On landing pages, swap generic product blurbs and corporate copy for answers to specific pain points.

Retailers, fill in Merchant Center’s conversational attributes, six optional product fields built for chatty shopper questions. And cut post-click friction, since humans take the easiest path.

Your customers now search in full sentences. Your ads and landing pages should answer in kind.

Advertisers want new audiences, but 70% would spend on the ones they already have

Got an extra ad budget burning a hole in your pocket?

Realize asked enterprise B2C advertisers what their main goal would be for more paid media money. The answers came out close to a four-way tie.

Reaching new audiences won, but barely, at 30%.

It’s the only option that looks beyond the people these brands already target. Add up the other three, and 70% would spend the extra money on familiar audiences.

That 70% splits almost evenly:

• 24% would show ads more often
• 24% would test new creative or messaging
• 22% would bid more aggressively on the same audiences

Just 8 points separate first place from last, so nobody agrees on a playbook.

Existing audiences feel safe, since the data is already there and results are easy to measure.

But more spend on the same people usually brings diminishing returns, and higher bids mostly raise your own costs.

What can you do: If your cost per acquisition rises as frequency climbs, your core audience is saturated. Move part of the new budget into prospecting and measure it against a holdout group.