Why Meta and Google now pick your ads from the material you supply
Meta and Google now decide who sees your ads and what search ads say, using the material you give them. Here is what that means for ads, press and one team.
Want this done for you? See our Google Ads management service. Related reading: Brand building vs sales ads: how to split a budget · How Google's ad auction prices relevance, not just bids · Branded searches: from ad view to purchase.
Key takeaways
- Meta narrows tens of millions of ad candidates to a few thousand before ranking, so an ad has to be retrieved before it can be ranked.
- Google's AI Max generates ad copy and chooses landing pages from your site, so the quality of that material shapes the result.
- In Blue Leopard Media's view, press coverage gives both platforms more distinct, credible material to work with, so ads and press should run as one team.
In short
Meta's Advantage+ automates budget allocation, audience targeting and bid adjustments. Google's AI Max for Search campaigns widens search term matching and generates ad copy.
The platform still needs something to work from. Meta chooses from the ads it is given, and Google's AI Max reads the keywords, ads and pages a business supplies. That changes what a founder should be buying, and it explains why Meta Ads, Google Ads and press coverage work better together than apart.
Meta starts with a cull, and your ads have to survive it
Meta's engineering team describes retrieval as the first stage of its ads system. It narrows "tens of millions of ad candidates" to a few thousand. Later ranking models then choose the final ads a person sees.
Meta's name for the retrieval system is Andromeda, announced on December 2, 2024. Meta reports a 6% recall improvement to retrieval and an 8% ads quality improvement on selected segments, with model capacity up 10,000 times.
The practical point is where the contest is decided. An ad that is not retrieved is never ranked.
Meta also says the Andromeda model architecture is expected to deliver a more diverse set of ad candidates, and states that "Increased ad diversity can improve people's experience with ads and drive better advertiser outcomes."
The same post says the number of ad creatives is expected to grow significantly as generative AI tools are adopted. Meta reports that over a million advertisers created more than 15 million ads in a month with those tools, and that Andromeda uses a hierarchical index to cope with that growth. The pool of ads competing to be retrieved is getting larger.
Google's AI Max reads your website to decide what to say
Google's Ads Help page lists search term matching, text customization and Final URL expansion among the features of AI Max for Search campaigns. Search term matching reaches beyond existing keywords using broad match and keywordless technology. Text customization generates ad copy for specific searches. Final URL expansion sends traffic to the pages on your site that Google judges most relevant.
The inputs are your own. Search term matching learns from current keywords, creatives and URLs. Text customization draws on existing ads, landing page copy and assets, plus generative AI. Final URL expansion chooses from the URLs on your website, and it requires text customization to be on.
There are trade-offs. With Final URL expansion on, pinned assets are not used if a more relevant URL is chosen. Google also warns that AI Max is not effective if a campaign is limited by budget. Advertisers keep some controls, including brand inclusions and exclusions, URL inclusions and exclusions, and a toggle for search term matching at ad group level.
Google announced AI Max on May 6, 2025, as a beta for all advertisers globally from later that month. It said advertisers who activate it will typically see 14% more conversions or conversion value at a similar cost per acquisition or return. For campaigns still mostly using exact and phrase keywords, Google cites 27%, based on campaigns with more than 70% of conversions or conversion value from those keywords for non-Retail advertisers. Those are Google's own figures from internal data, and they are not a promise about any one account.
What each channel misses on its own
Meta's retrieval engine is built for personalization. Meta describes the 10,000-times increase in model capacity as being for enhanced personalization. Meta says it uses machine learning to predict which ads a person will find most interesting.
Google Ads catch people who are searching. Google's own system now expands beyond exact keywords, so the ad can appear on searches the advertiser did not list.
In Blue Leopard Media's view, each channel fails in a different place. Meta finds without convincing, Google catches without proving, and press proves without reaching.
What press coverage adds to both platforms
This part is Blue Leopard Media's reading of the platform documents above, not a claim from either company. Google's AI Max draws on the keywords, creatives, landing pages and URLs a business supplies, and Meta's system draws on the ads it is given.
For Meta, a feature gives a new angle to build ads around: what a publication chose to say, which differs from the business's own pitch. That may widen the set of distinct ads available for retrieval, which is the kind of diversity Meta says it expects.
For Google, a feature gives a searcher a place to check the claim after clicking. It also gives the business material to quote and link to from its own pages, which is the landing page copy that text customization draws on.
None of this guarantees a particular result, and Blue Leopard Media does not promise one. It is a description of why the same budget has more to work with when press sits behind the ads.
What running them together looks like at Blue Leopard Media
Blue Leopard Media, founded in 2020, is a small team working with clients in the US, UK, Canada and Australia. It runs Google Ads and Meta Ads by custom proposal, makes changes weekly and reports monthly. It also writes and places editorial features on the publications it works with.
Features cost $249 for one publication, $499 for three and $999 for eight, paid once with no retainer. The client approves every word before anything is published.
In Blue Leopard Media's view, Meta Ads, Google Ads and press coverage work better as one team, so the founder does not have to carry material between vendors. The feature informs the ads, and the ads send people to a business that already has coverage behind it.
A founder can apply a simple test to any provider. Ask what the platform will read and choose from, and who is responsible for making that material good. If the answer is that it sits with someone else, the platforms' automation will work with whatever that someone else produced.
Sources
- Meta Engineering: Meta Andromeda: Supercharging Advantage+ automation with the next-gen personalized ads retrieval engine
- Google Ads Help: Optimize Google Ads with AI Max for Search campaigns
- Google: Unlock next-level performance with AI Max for Search campaigns
- Blue Leopard Media: About Blue Leopard Media
- Blue Leopard Media: PR features
Keep reading.
Why most of a budget should build the brand, not only chase sales
IPA data points to just over 60% on brand building and under 40% on activation. See how Meta Ads, Google Ads and press cover each side of that split.
Read the guide Ads + pressHow Google's ad auction prices relevance, and why a higher bid is not the only lever
Google ranks ads on quality and context as well as bid. See what its auction scores, how higher quality can lower costs per click, and where press coverage fits.
Read the guide Ads + pressBranded searches: the step between seeing an ad and buying
Google counts searches for your brand after a YouTube or Demand Gen ad view. See what the metric misses and why your results page matters.
Read the guideAds and press, run by one team.
Google Ads, Meta Ads and press features planned together, changed weekly and reported monthly. Tell us what you sell and we will send a proposal.