Daily Marketing News, October 13, 2026: Google Ads tests building campaigns from plain text
Google Ads is checking whether you can launch an ad campaign just by writing out in words what you want to achieve. In YouTube reach advertising, Google now lets you skip entering a bid, and Microsoft quietly put a model in Bing that grades the answers of other AI models.
Story of the day: Google Ads tests launching a campaign from a goal described in plain text
On October 12, 2026 it became known that Google Ads, Google's ad platform for Search, YouTube and partner sites, is testing a new way to create campaigns. Until now a campaign started with choosing a ready made goal from a list and a campaign type. In the test, the first screen asks "What's your goal for this campaign?", and below it is a large field where you write the goal in your own words.
Then Google's AI takes over. It reads that text and suggests campaign goals and assets, meaning headlines, descriptions, images and other ad elements. The old path hasn't been removed: the interface still has a "Create campaign manually" button that leads to the familiar manual setup, where the AI suggests nothing.
The test was confirmed by Ginny Marvin, Google Ads Liaison, reports Search Engine Land. According to her, it's an experiment, and Google wants feedback on where this flow would be useful. ContentGrip adds a reminder: everything the AI suggests still has to be checked by the advertiser before launch.
Why this matters to anyone paying for ads. The interface is moving from checkboxes to text, and a campaign now turns out exactly as good as the brief is precise. Write "I want more sales" and you get ads aimed at everyone and settings picked blindly. Write what you sell, to whom, in which city, at what average order value and what counts as a result for you, and you get a shot at a campaign you won't have to rebuild.
What to do right now, even if you don't have the test. Put a short brief in one document: the product, who buys it, what result you need, the budget, the geography and what's off limits. That text will be useful here and in any other ad platform that moves setup to free text. And read the assets Google generates with your own eyes before launch: a mistake in an ad headline costs money from the very first impression.
More news for October 13, 2026
Video Reach campaigns are YouTube reach campaigns: the advertiser pays for every 1,000 impressions, a model known as CPM. Until now, when creating a campaign you had to manually enter a Target CPM, the price you want to pay per 1,000 impressions. Now that field can be left empty, and Google changes bids itself over the whole campaign period so the budget is spent to 100%, according to Google's help center. It works in 2 subtypes: Efficient reach and non skippable campaigns with Target CPM. During the beta the feature is available only to eligible accounts, and everyone else still sets the bid by hand. Running reach ads will get easier, but the algorithm's job is to spend the budget, not to buy impressions cheaply, so the price of 1,000 impressions will be in Google's hands. If you have a hard CPM ceiling, I'd keep the manual bid and test the automation on part of the budget.
Muse is Meta's family of models and also the name of its personal AI agent. Since its US launch in September, the agent has run on the simpler Muse Spark model. Wonder is already open to developers through the Meta Model API and in Muse Code, Meta's coding service, while Wonder Full is available only on that service's Power Usage and Maximum Usage plans. Both models handle up to 1 million tokens of text at once and remember important details between conversations. In Meta's internal memory test, Wonder scored 85.6%, Claude Opus 5.5 scored 81.5% and GPT-6 Astra scored 81.4%. The agent already has an audience: Muse launched on September 8 and reached 5 million downloads in 22 days, reports SiliconANGLE. For brands, this is one more assistant answering people's questions about products and services, and it's worth being visible in its answers too.
Bing is Microsoft's search engine, and Copilot is its AI assistant. The Decision-1 model doesn't write text. It receives a situation, a question and a fixed set of answer options and returns a probability for each option, which means it sorts queries, classifies them by type and checks the answers of other AI models. It's built on the open model Qwen3.5-9B and costs $0.042 per 1 million input tokens, with output tokens free. According to Microsoft, on review analysis its quality is comparable to GPT-6 Sol, while it runs more than 14 times faster and costs 200 times less. Bing isn't applying the model to every query or every user, but the experiment is live. For anyone getting traffic from Bing, this is one more invisible filter deciding what kind of query it is and whether an answer is good enough.
This is about rank trackers and SEO platforms: they send queries to Google themselves and read the results page with scrapers, programs that automatically collect data from web pages. For several weeks now Google has been blocking these programs more successfully than it has in a long time. The publication wrote about this a few weeks ago and is surprised that many tools still haven't adapted. For site owners and SEO specialists, this means rankings in third party trackers may currently be incomplete or simply wrong. A sharp drop in a tracker report is not necessarily a drop on your site. Before you change anything, check the numbers against Search Console, Google's free dashboard for site owners that shows real clicks and impressions.
AI Overviews are the AI answers Google places above the regular results. It turns out the AI answer often appears when search can't find relevant pages or decides the query is a typo. In those cases the box may say the message looks like a typo or random keystrokes. Barry Schwartz, the publication's editor, noticed something else too: AI Overviews now even point to Google's own old search tools and widgets. In practice the AI box has become an almost mandatory part of the results page, and a person may never scroll down to the links below it. Site owners should search Google for their brand name with typos and for unusual phrasings of their services, and see what the AI answer tells people instead of sending them to your site.
This is about products in Merchant Center, the Google platform where stores upload their catalog for Google Shopping and product ads. Search Engine Roundtable relayed the statement in its October 12 recap: a poor photo on its own won't stop a product from showing in Google Search or in ad campaigns. For online store owners, this means a weak image is not a reason for a listing to be disapproved and not the reason a product disappeared from results. Look for that reason elsewhere: price, availability, feed data. But people still choose with their eyes, and the fact that Google let a blurry photo through doesn't mean anyone will click it and buy.
The report's main point, according to BestMediaInfo, is that a low acquisition cost and a strong campaign response can hide poor customer quality, weak activation and little long term value for the business. The piece ran under the headline "a click is not a customer". The report is about India's financial market, but the idea works for anyone paying for ads: if you only count cost per click or per lead, your campaigns gradually learn to bring in cheap people who buy nothing. It's easy to check: take last month's leads, see how many of them got to payment, and which channel brought the people who bought again.
What advertisers, creators and site owners should do about it
- Google Ads advertisers: write a short text brief of 5 lines: the product, who buys it, what result you need, budget and geography, and what's off limits. Ad platform setup is shifting more and more to text, and the goals and ads the AI suggests depend on how precise your brief is. Check everything it suggests by hand before launch.
- Anyone buying reach on YouTube: try automatic bidding in Video Reach on part of your budget and compare actual CPM, reach and frequency against a campaign on manual Target CPM.
- Site owners: while third party trackers are failing, check your rankings against Search Console. Separately, search Google for your brand name with typos and see what the AI answer above the results says.
Writes about: SMM, paid boosts, ads, traffic, suppliers, the market from the inside, daily news
