A Real Google Ads Audit Walkthrough: Brand vs Non-Brand Shopping, Impression Share, PMax and Demand Gen

I recently ran a full audit on a lovely little ecommerce brand, and I want to walk you through exactly what I found and exactly what I told them to do about it. It is one of those accounts that looks perfectly healthy on the surface, yet has tens of thousands of dollars of easy wins sitting there untouched.

The brand sells laundry-care consumables in the Netherlands. Low ticket, repeat purchase, a small catalogue of around ten products, and a lovely high conversion volume because people keep coming back for more. They were already running Standard Shopping and doing well on Meta with cold prospecting, but there was no Performance Max history and nothing at the top of the funnel on Google. Their goal was modest: a 2x ROAS on new customers.

Here is everything I picked apart.

The Brand vs Non-Brand Split Nobody Was Controlling

The single biggest issue was that all their Shopping traffic ran through one campaign. Branded and non-branded searches were mixed together, which means you cannot control them independently, and they behave nothing alike.

When I broke the numbers down, branded terms accounted for only around 22% of Shopping spend but almost 50% of the revenue. Non-brand was the opposite: roughly 80% of spend for about 55% of revenue. Two completely different animals living in one cage.

My recommendation was to segment them into separate campaigns so each can have its own ROAS target. I suggested aiming for roughly 6.5x on brand and around 3x on non-brand. And here is a nuance I really want you to sit with: brand and non-brand are not the same thing as returning versus new customers. Non-brand still contains plenty of returning buyers, and brand contains genuinely new customers who researched elsewhere and then searched the name. So set your targets on the actual economics, not on a lazy new-versus-returning assumption. That is also why their 2x non-brand goal was actually too low once you strip out the returning buyers hiding in there.

The way to do this with Standard Shopping is refreshingly simple. Take the existing campaign, treat it as your non-brand campaign, and add all your brand terms as negative keywords so branded queries can no longer sneak in. Then launch a second Standard Shopping campaign without those negatives, so branded queries flow into it instead. Set the brand campaign to High priority and the non-brand campaign to Low priority to reinforce the split. It is never perfect, so expect around 1 to 2% leakage even when maintained well, but it works. One warning: this priority trick only exists in Standard Shopping. You cannot do it if non-brand runs as Performance Max.

There is also a native brand inclusion and exclusion setting rolling out worldwide in beta. Once it lands, you set the brand as an inclusion in the brand campaign and an exclusion in the non-brand one, which is far cleaner than the negative keyword workaround. With this account's healthy conversion volume, it is a strong candidate to test.

The Free Impression Share Sitting on the Table

This one genuinely excites me. The campaign was losing around 37% impression share to budget, sitting at roughly 62% impression share overall.

That word, budget, is the important part. When you lose impression share to budget rather than rank, raising the daily budget lifts your market share directly. No bid changes needed. You are simply choosing to be visible more often at a very similar ROAS.

There was a ticking clock too. The campaign ran a 2.5x ROAS target but was over-delivering at about 2.7x, purely because budget-capping forces Google to lower CPCs and inflate ROAS. That over-delivery behaviour ends on 17 August. After that, a budget-capped campaign simply delivers the target, not the inflated version. So my advice was to act decisively before that date: roughly double or even triple the budget in one move. That shifts the budget-lost impression share over to rank-lost, jumps market share by around 36 to 37%, and turns the campaign into a ROAS-limited one, which is exactly where you want to be.

One simple rule of thumb to remember: the goal should be that impression share should only ever be lost to rank, never to budget. If you are losing impression share due to budget and you are at a ROAS target you are happy with, then you can either raise your budget and get more traffic at the target ROAS you set, or you can raise your ROAS target and get higher ROAS for the same spend. Either of those two options is better than continuing to lose impression share due to budget on a campaign that has a ROAS target you are happy with.

Two Feed Fixes Worth Real Money

Their product titles were nicely optimised in Dutch, but the product type field was completely empty. That is a mistake, because product type is the second highest-leverage feed field after the title for telling Google which keywords to show your Shopping ads against. You can go up to five levels deep with a 750 character limit, so I told them to fill every level with relevant keywords, using the same principles as their titles.

The second fix is a quiet one. Because they advertise in the EU, using a Comparison Shopping Service provider hands you an automatic discount of around 20% on every single click, thanks to the EU antitrust ruling. You cannot verify from inside the ads account whether one is active, so I flagged it to confirm. If it is missing, they are paying 20% more per click than they need to.

Testing Performance Max Properly

With no Performance Max history at all, I recommended a test. PMax often runs at a slightly lower ROAS than Standard Shopping but usually pulls more total volume, because it bolts retargeting and search text ads onto the Shopping component. More volume at a solid ROAS means more profit.

The key is giving it a fair test: at least 500 conversions in PMax versus 500 in Standard Shopping over a similar period. If you are wary of the two overlapping and stealing from each other, run a proper campaign experiment instead, a scientific A/B split where audience A never sees B's ads. Google then tells you whether PMax produced a statistically significant improvement. When you do run it, go full assets: reuse the best Meta video creatives and top images, and with only ten products, use one asset group per product so you can tailor headlines, images and descriptions right down at the product level.

Extending Their Meta Success to Google's Top of Funnel

Every scrap of their Google spend was bottom of funnel, capturing existing demand. There was nothing generating fresh awareness, despite the fact that Meta prospecting was already working beautifully for them. That same creative should work top of funnel on Google through Demand Gen.

The measurement piece matters enormously here. You must enable platform comparable conversions, which measures Demand Gen more like Meta measures its own ROAS, and then benchmark against your Meta ROAS rather than your normal Google target. Judging top-of-funnel by a bottom-of-funnel yardstick is how people kill campaigns that were actually working. Alongside that, watch your total Google Ads revenue, branded search volume and overall bottom-line revenue, because that is where the awareness effect really shows up.

Budget and patience are non-negotiable. At $500 a month the impact is too small to even detect, so given their low AOV and market size I recommended at least $5,000 to $10,000 a month, run for a minimum of two months and ideally three. Demand Gen needs frequency to build, because people react after around ten exposures, not one or two. Do not judge it in month one. For audiences, I suggested two lookalikes, a 5% of all customers and a 5% of high-value customers, plus their best in-market audiences, and a broad targeting test as well. Keep the structure simple with one prospecting and one retargeting campaign, run Maximise Conversions indefinitely, and only test Target CPA or Target ROAS once results look strong around month three or four.

That is the whole audit. A healthy account on the surface, and a small fortune in easy wins underneath.

Conclusion

A quick recap of the main points from this audit:

• Segment brand vs non-brand Shopping. Brand was around 22% of spend but nearly 50% of revenue, so split it into separate campaigns with separate targets (roughly 6.5x brand, 3x non-brand).

• Do not confuse brand/non-brand with returning/new. Set ROAS targets on economics, not on a false assumption.

• Claim your budget-lost impression share. Around 37% was lost to budget, not rank, so raising budget lifts market share at a near-identical ROAS.

• Beat the 11 August bid-strategy change. The 2.5x target over-delivering at 2.7x ends then, so raise budget and reset the target to what you actually want first.

• Fill the empty product type field. Five levels, 750 characters, keyword-rich. It is the second most important feed field after the title.

• Claim the EU CSS discount. A Comparison Shopping Service provider is around 20% off every click. Confirm one is active.

• Test PMax properly. Around 500 conversions each versus Standard Shopping, or an uplift experiment if wary, with full assets and one asset group per product.

• Extend Meta success to Google's top of funnel. Demand Gen with the best Meta creatives, $5,000 to $10,000 a month, a two to three month minimum, judged on platform comparable ROAS plus branded search and total revenue.

AI Max for Shopping: What It Really Does to Standard Shopping Campaigns

Google has just bolted AI Max onto your Shopping campaigns, and everyone online is fixated on the two features Google is loudly promoting. The problem is that both are almost sideshows. The feature that actually matters is the one Google is barely mentioning, and it might change how you structure your whole account. I run real ecommerce accounts every day through my agency, Big Flare, so let me give you the straight story rather than the noise.

First, You Might Not Even Have It Yet

AI Max for Shopping is currently in a closed beta, announced on 30 April 2026, so it has only landed for some accounts so far. I have checked my own client campaigns and do not see it active yet, so you might have it or you might not. There is no public opt-in form. If you want it early you would typically need a Google rep, but either way it will roll out to everyone eventually.

To be clear, this is my read of Google's own documentation and research, not a data-backed verdict. My team will run this beta extensively across accounts, and I will follow up with the tested strategy and real numbers. Treat everything here as a strong hypothesis.

What AI Max for Shopping Actually Does

At its core, it bolts a set of AI-powered features onto your Standard Shopping campaigns. Google is heavily touting two of them.

Feature One: Text Customisation

This lets Google rewrite or add to your product ad title depending on what the person actually searched for. It pulls information from your Merchant Center feed and tweaks the title to make it more relevant to that specific query. It is also built to perform in the newer search experiences like AI Overviews and AI Mode.

Feature Two: Final URL Expansion

This lets Google override the landing page you set and send the click to whatever page on your site it thinks will convert best instead.

The Feature Google Isn't Talking About

Here is the one that matters. When you turn AI Max for Shopping on, your Shopping campaign can now decide to show search text ads instead of only Shopping ads, served straight out of your Shopping campaign and built off your landing pages and product feed.

Those two headline features are what unlock it. In Google's own words, Final URL expansion advances your eligibility beyond basic Shopping ads so the campaign can serve hybrid text and product formats, and Google is clear that Final URL expansion needs Text Customisation switched on alongside it. So the real reason those two features matter has nothing to do with your Shopping ads; together they unlock search ads out of your Shopping campaign. On top of that, you also improve your distribution into AI placements like AI Overviews and AI Mode.

Why the Two Headline Features Are Minor for Your Shopping Ads

Both features are fine and will help a little, but for your Shopping ads themselves the impact is small.

Text Customisation matches your product title to the search, but the title just does not have a huge effect on whether someone clicks. What drives the click in Shopping is the image, the price, and the annotations like strikethrough pricing, a sale badge, and free shipping. That is where people's eyes go. The title's real job is not the click at all; it is targeting, which searches you show up for in the first place. So tweaking it to match the query makes a relatively small difference.

Final URL expansion barely moves the needle for Shopping either. The default landing page is already the best one, the exact product page for the exact product the person searched for, and it is very hard for Google to find a better page than that. Google's own documentation touts up to a 5% improvement in conversions here, and honestly, 5% is not a huge deal, especially since Google usually overvalues these numbers. So both heavily-promoted features add up to minimal impact on the Shopping ads themselves, which is exactly why I want to redirect your attention.

The Real Story: Search Text Ads on Standard Shopping

The best way to understand this is to compare it to Performance Max. Run PMax for ecommerce and Google places your ads across a whole spread of placements: YouTube, display, search text ads, Shopping ads, Gmail, and maps. What we almost always find is that the best-performing placement is Shopping ads, and the second-best is search text ads. Both normally perform really well at a good ROAS. After that, it is often a coin toss whether display, YouTube, Gmail, maps and the rest perform at all. So the two placements you actually want are Shopping and search text ads. This is exactly why we sometimes run a PMax feed-only campaign, which strips out all your assets to focus on Shopping ads only.

My long-standing bugbear with Performance Max is that it forces an all-or-nothing choice on your assets. Either you go full assets across all placements, or no assets for Shopping only. Google will not let you run search assets alone, so you cannot say, “I just want Shopping plus search”, which are the two best things. My workaround for years has been to run feed-only PMax for the Shopping side, then build out my search campaigns really well so I still capture that valuable search traffic separately.

Here is what is genuinely interesting about AI Max for Shopping: in theory, it gives you the best stuff only. Your Standard Shopping campaign still focuses purely on Shopping, so it does not suddenly start showing on Gmail, maps, display, or YouTube. But with this one button ticked, you can now pull search text ads into it as well. That is the two best-performing ecommerce placements, in one campaign, prioritised, without the all-or-nothing PMax asset trap.

Why You Actually Want Both Headline Features On

Remember those two headline features I called minor? I actually think you will want both on, not because of what they do for your Shopping ads, but because of what they do for the text ads.

Your Shopping ads do not really need Final URL expansion, but the text ads absolutely do: the system needs it to find the best page on your site to send that click to. And Text Customisation is what tailors the ad copy to match the query for those text ads. So the features Google frames as improving your Shopping ads do their real work powering the text-ad expansion. That is why you switch them on.

The Guardrails You Need to Set First

Once you have auto-generated text ads in the mix, you need guardrails before you launch.

Text guidelines are now, in my view, a required part of this strategy. They let you give the system natural-language direction on what you do and do not want in your copy, just like in AI Max for Search. Because the system is genuinely generating text ads here, that guidance really matters.

URL exclusions pair beautifully with Final URL expansion. They let you tell the system which pages to stay away from, such as your contact page, privacy policy, and possibly your blog, keeping it focused on your higher-intent, converting pages.

Brand settings are a lovely addition, letting you set brand inclusions and exclusions so segmenting brand versus non-brand in Shopping is far easier. We will probably still run our usual higher-priority non-brand campaign with the brand terms negated, but now we have an extra lever: on your non-brand campaign, exclude your brand; on your brand campaign, include only your brand.

The New Reporting Is Where the Work Happens

The reporting has genuinely improved, with more retail-centric data and a brand-new Asset report. Inside it, a tab called Expanded final URL assets shows how Text Customisation is performing: the text Google is creating for you and which landing pages it is sending traffic to. Crucially, you can remove assets, so anything off-brand or underperforming can be pruned.

This is not a fire-and-forget toggle. Set your guardrails, launch, then actively watch that reporting and prune what the system generates; the bigger your account and spend, the more that attention is worth. And remember, all of this leans on your feed and landing pages, so garbage in, garbage out now extends to your landing pages too. A clean feed and good landing pages are the foundation.

A New Fourth Option That Might Reshape Your Account

For years we had three choices: Standard Shopping, PMax full-assets, or PMax feed-only. Now there is a fourth: Standard Shopping with AI Max for Shopping bolted on.

I think this fourth choice might start taking priority over PMax feed-only. Feed-only PMax focuses on Shopping only, but what we have always actually wanted is Shopping plus search, and Standard Shopping with AI Max can now do exactly that. This might genuinely reduce the need for feed-only PMax.

Again, treat this as a strong hypothesis, not a verdict. I need real data first, and the tested follow-up is coming once we have run this across accounts.

If you already have beta access, follow the order above: text guidelines first, then Text Customisation and Final URL expansion, then brand settings and URL exclusions, then monitor that Asset report and prune anything that does not fit. If you do not have it yet, which is most of us, simply be ready: sort out your feed and landing pages, and start drafting your text guidelines now.

Conclusion

AI Max for Shopping bolts AI-powered features onto Standard Shopping. Google is promoting two of them, Text Customisation and Final URL expansion, but both have only a minor effect on your Shopping ads. The feature that really matters is the one Google is barely mentioning: with AI Max on, your Standard Shopping campaign can now serve search text ads, giving you the two best-performing ecommerce placements, Shopping plus search, in one campaign without Performance Max's all-or-nothing asset trap. You will still want both headline features on, because they power those text ads. Set your guardrails first, text guidelines, URL exclusions and brand settings, then launch and actively prune using the new Asset report. This creates a genuine fourth account structure option that may reduce the need for feed-only PMax, though that is a strong hypothesis pending real data.

August 2026 PPC Roundup: ChatGPT Ads, Google's Smart Bidding Change, Amazon's Title Cap and Meta's New Creative Rules

There has been an unusual amount of movement across ad platforms over the past month, and a few of these changes carry hard deadlines or automated penalties attached. So rather than let them slip past you, I want to walk through the five that actually matter for anyone running paid traffic on an ecommerce store right now. Here is what I would be checking this month.

1. OpenAI Has Built a Whole Ad Platform Inside ChatGPT

This is the one that genuinely made me sit up. Over a single month, OpenAI has quietly assembled a real, working ad platform inside ChatGPT, and it is maturing at a pace that is honestly a little alarming.

We are not talking about a vague "ads are coming" announcement. In the space of weeks they have shipped conversion-optimised bidding, custom audiences for remarketing, bid multipliers, website-pulled auto-generated creative, and now promotional codes, which is the classic "coupon stage" every ad business seems to pass through on its way to becoming serious. The most interesting piece is a new "Agent" ad format that launches a branded AI conversation instead of dropping the user on a landing page. Imagine your ad opening a chat that answers questions and guides someone to purchase, rather than a static page they bounce off.

Japan Airlines is already among the first big brands publicly building a dedicated ChatGPT ad strategy, and I won’t be surprised to see more brands hopping on board soon too. My advice for you is simple: start paying attention now and test small. The advertisers who learn a new channel early are always the ones who win cheap inventory before everyone else piles in.

2. Google's 17 August Smart Bidding Change Could Quietly Drop Your ROAS

This one has a date on it, so it is the most urgent item here. From 17 August, budget-limited Smart Bidding campaigns will start tracking much more closely to your set Target ROAS, instead of over-delivering above it the way many of them do today.

Here is why that matters. Right now, a budget-capped campaign often runs above its Target ROAS, because Google throttles your CPCs to make the spend fit inside the budget, which artificially inflates your returns. It feels lovely. You set a target and the campaign quietly beats it. From 17 August that over-delivery behaviour ends, and the campaign simply delivers the target you actually set, no more and no less.

So if you have campaigns cruising comfortably above their Target ROAS on a capped budget, that cushion is about to disappear. The fix is to be proactive: raise your Target ROAS ahead of the date to reflect the number you are genuinely happy with, rather than the inflated figure you have been quietly enjoying. If you do nothing, you risk waking up on 18 August to a noticeably lower ROAS across affected campaigns. Audit every budget-limited Target ROAS campaign now, and reset the targets before Google resets them for you.

3. Amazon Will Rewrite Your Product Titles If You Don't

Amazon has enforced a hard 75-character limit, spaces included, on product titles across every category except media. Anything beyond that now belongs in a new 125-character "Item Highlights" field instead.

The sting is in the enforcement. Sellers who do not manually update their non-compliant titles within 14 days will have them automatically replaced by Amazon's own AI-generated version. Read that again, because it is the important bit. If you miss the window, you hand control of your single most important piece of listing copy over to an algorithm that does not know your brand, your keywords, or which terms actually drive your sales.

If you sell on Amazon, this is a same-week job. Audit every title, trim the ones over 75 characters yourself, and move the detail you care about into Item Highlights while you still control the wording. Do not let Amazon's AI make those calls for you.

4. Google Ads Now Tells You How Much Revenue You're Missing

On a cheerier note, Google has added "Missed Opportunity Reporting" to the Recommendations tab. It gives you modelled estimates of the clicks, conversions and revenue you are losing to under-spending or overly cautious bidding, along with suggested actions such as raising a budget or lowering a ROAS target.

I like this as a diagnostic, because it puts a number on the "we are leaving money on the table" conversation that used to be all hand-waving. That said, treat these figures as directional rather than gospel. They are Google's own models, and Google has an obvious incentive to encourage more spend, so validate anything eye-catching against your own account data and margins before you act. Used with a healthy dose of scepticism, it is a genuinely useful starting point for spotting where a capped budget or a too-conservative target is holding you back.

5. Meta Just Changed What Counts as "Different" Creative

Finally, a more forward-looking one that changes how you should think about creative testing. Meta published research on a new upstream ad-ranking layer called Hierarchical Interest Representation. In plain English, it now builds a unified understanding of your ad from the actual content of the creative itself, not just who made it or who it is shown to.

The practical implication is the part worth sitting with. If your "different" variants only swap the actor or the presenter while keeping the same underlying pitch, Meta's algorithm may increasingly register them as the same ad, rather than genuinely differentiated creative. That is a big shift. For years, changing the talent counted as a new test. Going forward, real differentiation needs to come from different buyer motivations, different problem framings, or different angles on the offer, not just a fresh face reading the same script.

So if you have been running lots of near-identical variants and wondering why you are not seeing incremental lift, this is likely part of the answer. Build your next round of creative around genuinely distinct messages, and you will give the algorithm something real to work with.

That is the month's most important movement in one place. A new channel worth watching, a bidding change with a hard deadline, an Amazon compliance clock ticking, a handy new Google diagnostic, and a rethink of what creative testing even means on Meta. Pick the ones that touch your accounts and action them this week.

Conclusion

A quick recap of this month's key changes:

• ChatGPT Ads are becoming real. In one month OpenAI has shipped conversion bidding, custom audiences, bid multipliers, auto-generated creative, promo codes and an "Agent" ad format that launches a branded AI chat. Start watching this channel and start testing small now.

• Google's 17 August Smart Bidding change. Budget-limited campaigns will track closer to your Target ROAS instead of over-delivering above it. Raise your targets to what you genuinely want before the date, or risk an overnight ROAS drop.

• Amazon's 75-character title cap. Titles over the limit are auto-replaced by Amazon's AI within 14 days. Audit and trim your own titles now, and move extra detail into the new Item Highlights field.

• Google's Missed Opportunity Reporting. New Recommendations-tab estimates of lost clicks, conversions and revenue. Useful and directional, but validate against your own data and margins before acting.

• Meta's new creative ranking. Variants that only swap talent may now register as the same ad. Real differentiation needs different motivations and problem framings, not just a different face.