- Between August 5 and 7, 2026, Google AI Overviews on “best/top + location” searches fell from a steady 70.7% to 29.5% — a 41-point drop.
- It hit every business, industry and market we track, in the same three-day window. That rules out any one website being the cause.
- The trigger is local intent, not the word “best.” National “best of” searches barely moved (94.5% → 70.5%); location-based searches without a superlative fell nearly as hard as the main group.
- It may be a response to the AEO/GEO optimization wave, a permanent re-scoping of local results, or a lull before a bigger Google update. We can’t confirm which.
- The practical lesson: without daily measurement you’d read this as your failure. It wasn’t.
If you searched “best BBQ in Kansas City” or “top real estate agents near me” back in July, Google almost certainly answered you with an AI Overview — the AI-written summary that sits above every other result. Search the same thing today, and there’s a good chance that AI answer is simply gone.
That’s not a hunch. It’s what our daily tracking data shows — and the shift happened fast, over about three days in early August.
What we measured
We continuously monitor how AI search engines answer real buyer questions for a portfolio of businesses. For this analysis we looked at more than 320 search prompts tracked daily across ten different businesses, spanning several industries — marketing services, retail shops, restaurants, and real estate, along with HR services, logistics, and home services — and several markets, from Midwest metros and suburbs to East Coast towns and South Florida, plus “near me”-style searches.
Every day, each prompt is run against Google and we record one simple fact: did Google show an AI Overview for this search, yes or no?
We focused on the searches that matter most to a local business’s bottom line — bottom-of-funnel (BOFU) queries where someone is actively choosing a provider: prompts containing “best” or “top” plus a location, like “best digital marketing agency in Lenexa, KS,” “top bridal shops in St. Louis,” or “who serves the best BBQ ribs in Kansas City?”
The drop: 71% to 30% in a matter of days
From late June through July, Google showed an AI Overview on 70.7% of our daily “best/top + location” prompt checks. The rate was remarkably stable — hovering between the mid-60s and low 80s for six straight weeks, peaking at 82% on July 24.
Then, between August 5 and August 7, it collapsed. Since August 7, the same prompts have averaged just 29.5% — a 41-point drop. The single lowest day was August 11, at 19.7%.
View data as table
| Date | Prompts checked | AIO shown | No AIO | % with AIO |
|---|---|---|---|---|
| 2026-06-22 | 124 | 86 | 38 | 69.4% |
| 2026-06-23 | 127 | 78 | 49 | 61.4% |
| 2026-06-24 | 128 | 83 | 45 | 64.8% |
| 2026-06-25 | 128 | 86 | 42 | 67.2% |
| 2026-06-26 | 128 | 87 | 41 | 68.0% |
| 2026-06-27 | 128 | 87 | 41 | 68.0% |
| 2026-06-28 | 128 | 86 | 42 | 67.2% |
| 2026-06-29 | 128 | 85 | 43 | 66.4% |
| 2026-06-30 | 128 | 97 | 31 | 75.8% |
| 2026-07-01 | 128 | 86 | 42 | 67.2% |
| 2026-07-02 | 128 | 79 | 49 | 61.7% |
| 2026-07-03 | 128 | 93 | 35 | 72.7% |
| 2026-07-04 | 128 | 85 | 43 | 66.4% |
| 2026-07-05 | 128 | 86 | 42 | 67.2% |
| 2026-07-06 | 128 | 89 | 39 | 69.5% |
| 2026-07-07 | 128 | 91 | 37 | 71.1% |
| 2026-07-08 | 128 | 92 | 36 | 71.9% |
| 2026-07-09 | 128 | 90 | 38 | 70.3% |
| 2026-07-10 | 128 | 88 | 40 | 68.8% |
| 2026-07-11 | 128 | 94 | 34 | 73.4% |
| 2026-07-12 | 128 | 89 | 39 | 69.5% |
| 2026-07-13 | 128 | 92 | 36 | 71.9% |
| 2026-07-14 | 128 | 93 | 35 | 72.7% |
| 2026-07-15 | 128 | 95 | 33 | 74.2% |
| 2026-07-16 | 128 | 93 | 35 | 72.7% |
| 2026-07-17 | 128 | 97 | 31 | 75.8% |
| 2026-07-18 | 128 | 94 | 34 | 73.4% |
| 2026-07-19 | 128 | 96 | 32 | 75.0% |
| 2026-07-20 | 128 | 97 | 31 | 75.8% |
| 2026-07-21 | 128 | 93 | 35 | 72.7% |
| 2026-07-22 | 128 | 89 | 39 | 69.5% |
| 2026-07-23 | 128 | 93 | 35 | 72.7% |
| 2026-07-24 | 128 | 105 | 23 | 82.0% |
| 2026-07-25 | 128 | 89 | 39 | 69.5% |
| 2026-07-26 | 128 | 90 | 38 | 70.3% |
| 2026-07-27 | 128 | 89 | 39 | 69.5% |
| 2026-07-28 | 128 | 98 | 30 | 76.6% |
| 2026-07-29 | 128 | 93 | 35 | 72.7% |
| 2026-07-30 | 128 | 93 | 35 | 72.7% |
| 2026-07-31 | 128 | 92 | 36 | 71.9% |
| 2026-08-01 | 128 | 93 | 35 | 72.7% |
| 2026-08-02 | 128 | 93 | 35 | 72.7% |
| 2026-08-03 | 128 | 96 | 32 | 75.0% |
| 2026-08-04 | 128 | 74 | 54 | 57.8% |
| 2026-08-05 | 128 | 57 | 71 | 44.5% |
| 2026-08-06 | 128 | 61 | 67 | 47.7% |
| 2026-08-07 | 128 | 33 | 95 | 25.8% |
| 2026-08-08 | 128 | 41 | 87 | 32.0% |
| 2026-08-09 | 128 | 31 | 97 | 24.2% |
| 2026-08-10 | 137 | 37 | 100 | 27.0% |
| 2026-08-11 | 137 | 27 | 110 | 19.7% |
| 2026-08-12 | 137 | 43 | 94 | 31.4% |
| 2026-08-13 | 137 | 44 | 93 | 32.1% |
| 2026-08-14 | 137 | 51 | 86 | 37.2% |
| 2026-08-15 | 137 | 30 | 107 | 21.9% |
| 2026-08-16 | 137 | 55 | 82 | 40.1% |
| 2026-08-17 | 137 | 38 | 99 | 27.7% |
| 2026-08-18 | 137 | 40 | 97 | 29.2% |
| 2026-08-19 | 137 | 45 | 92 | 32.8% |
| 2026-08-20 | 137 | 43 | 94 | 31.4% |
This wasn’t one client, one industry, or one city. Every business we track — across every industry and every market in the dataset — dropped within the same roughly three-day window. When ten unrelated businesses in different industries and different cities all move together, the change isn’t about anyone’s website. It’s about Google.
It’s the location that triggers the pullback, not the word “best”
To isolate what changed, we split the rest of our tracked prompts into two comparison groups and pulled the same daily history:
- “Best/top” prompts with no location (e.g., “best PEO for small business,” “top biometric authentication vendors”) — these barely flinched. They ran at 94.5% before August and still average 70.5% after the shift.
- Location prompts without “best/top” (e.g., “where to buy a wedding dress in Kansas City,” “bridal shops in Fort Lauderdale with private suites”) — these fell almost as hard as the main group: from 84.0% down to 37.7%.
View data as table
| Date | best/top + location | best/top only | location only |
|---|---|---|---|
| 2026-06-22 | 69.4% (86/124) | 96.6% (57/59) | 82.4% (42/51) |
| 2026-06-23 | 61.4% (78/127) | 96.8% (60/62) | 86.3% (44/51) |
| 2026-06-24 | 64.8% (83/128) | 92.4% (61/66) | 78.4% (40/51) |
| 2026-06-25 | 67.2% (86/128) | 93.9% (62/66) | 78.4% (40/51) |
| 2026-06-26 | 68.0% (87/128) | 93.9% (62/66) | 82.4% (42/51) |
| 2026-06-27 | 68.0% (87/128) | 98.5% (65/66) | 80.4% (41/51) |
| 2026-06-28 | 67.2% (86/128) | 97.0% (64/66) | 82.4% (42/51) |
| 2026-06-29 | 66.4% (85/128) | 97.0% (64/66) | 88.2% (45/51) |
| 2026-06-30 | 75.8% (97/128) | 93.9% (62/66) | 78.4% (40/51) |
| 2026-07-01 | 67.2% (86/128) | 84.8% (56/66) | 74.5% (38/51) |
| 2026-07-02 | 61.7% (79/128) | 80.3% (53/66) | 68.6% (35/51) |
| 2026-07-03 | 72.7% (93/128) | 92.4% (61/66) | 84.3% (43/51) |
| 2026-07-04 | 66.4% (85/128) | 93.9% (62/66) | 84.3% (43/51) |
| 2026-07-05 | 67.2% (86/128) | 95.5% (63/66) | 84.3% (43/51) |
| 2026-07-06 | 69.5% (89/128) | 92.4% (61/66) | 78.4% (40/51) |
| 2026-07-07 | 71.1% (91/128) | 92.4% (61/66) | 80.4% (41/51) |
| 2026-07-08 | 71.9% (92/128) | 95.5% (63/66) | 84.3% (43/51) |
| 2026-07-09 | 70.3% (90/128) | 93.9% (62/66) | 88.2% (45/51) |
| 2026-07-10 | 68.8% (88/128) | 97.0% (64/66) | 86.3% (44/51) |
| 2026-07-11 | 73.4% (94/128) | 92.4% (61/66) | 86.3% (44/51) |
| 2026-07-12 | 69.5% (89/128) | 93.9% (62/66) | 82.4% (42/51) |
| 2026-07-13 | 71.9% (92/128) | 90.9% (60/66) | 82.4% (42/51) |
| 2026-07-14 | 72.7% (93/128) | 97.0% (64/66) | 84.3% (43/51) |
| 2026-07-15 | 74.2% (95/128) | 97.0% (64/66) | 84.3% (43/51) |
| 2026-07-16 | 72.7% (93/128) | 97.0% (64/66) | 86.3% (44/51) |
| 2026-07-17 | 75.8% (97/128) | 97.0% (64/66) | 88.2% (45/51) |
| 2026-07-18 | 73.4% (94/128) | 97.0% (64/66) | 90.2% (46/51) |
| 2026-07-19 | 75.0% (96/128) | 97.0% (64/66) | 86.3% (44/51) |
| 2026-07-20 | 75.8% (97/128) | 95.5% (63/66) | 88.2% (45/51) |
| 2026-07-21 | 72.7% (93/128) | 93.9% (62/66) | 82.4% (42/51) |
| 2026-07-22 | 69.5% (89/128) | 93.9% (62/66) | 82.4% (42/51) |
| 2026-07-23 | 72.7% (93/128) | 95.5% (63/66) | 82.4% (42/51) |
| 2026-07-24 | 82.0% (105/128) | 95.5% (63/66) | 98.0% (50/51) |
| 2026-07-25 | 69.5% (89/128) | 97.0% (64/66) | 82.4% (42/51) |
| 2026-07-26 | 70.3% (90/128) | 92.4% (61/66) | 84.3% (43/51) |
| 2026-07-27 | 69.5% (89/128) | 95.5% (63/66) | 86.3% (44/51) |
| 2026-07-28 | 76.6% (98/128) | 95.5% (63/66) | 90.2% (46/51) |
| 2026-07-29 | 72.7% (93/128) | 95.5% (63/66) | 86.3% (44/51) |
| 2026-07-30 | 72.7% (93/128) | 97.0% (64/66) | 88.2% (45/51) |
| 2026-07-31 | 71.9% (92/128) | 97.0% (64/66) | 86.3% (44/51) |
| 2026-08-01 | 72.7% (93/128) | 95.5% (63/66) | 84.3% (43/51) |
| 2026-08-02 | 72.7% (93/128) | 100.0% (66/66) | 84.3% (43/51) |
| 2026-08-03 | 75.0% (96/128) | 98.5% (65/66) | 86.3% (44/51) |
| 2026-08-04 | 57.8% (74/128) | 87.9% (58/66) | 64.7% (33/51) |
| 2026-08-05 | 44.5% (57/128) | 78.8% (52/66) | 54.9% (28/51) |
| 2026-08-06 | 47.7% (61/128) | 84.8% (56/66) | 70.6% (36/51) |
| 2026-08-07 | 25.8% (33/128) | 77.3% (51/66) | 37.3% (19/51) |
| 2026-08-08 | 32.0% (41/128) | 69.7% (46/66) | 37.3% (19/51) |
| 2026-08-09 | 24.2% (31/128) | 69.7% (46/66) | 39.2% (20/51) |
| 2026-08-10 | 26.6% (34/128) | 66.7% (44/66) | 27.5% (14/51) |
| 2026-08-11 | 19.5% (25/128) | 63.6% (42/66) | 29.4% (15/51) |
| 2026-08-12 | 32.0% (41/128) | 71.2% (47/66) | 47.1% (24/51) |
| 2026-08-13 | 32.0% (41/128) | 62.1% (41/66) | 41.2% (21/51) |
| 2026-08-14 | 38.3% (49/128) | 72.7% (48/66) | 39.2% (20/51) |
| 2026-08-15 | 21.9% (28/128) | 71.2% (47/66) | 43.1% (22/51) |
| 2026-08-16 | 40.6% (52/128) | 78.8% (52/66) | 35.3% (18/51) |
| 2026-08-17 | 27.3% (35/128) | 72.7% (48/66) | 37.3% (19/51) |
| 2026-08-18 | 28.1% (36/128) | 71.2% (47/66) | 37.3% (19/51) |
| 2026-08-19 | 32.8% (42/128) | 71.2% (47/66) | 41.2% (21/51) |
| 2026-08-20 | 32.0% (41/128) | 68.2% (45/66) | 35.3% (18/51) |
Put simply: Google didn’t back away from “best of” questions. It backed away from local questions. Any search with local intent — with or without a superlative — lost its AI Overview roughly half to two-thirds of the time, while national and industry-level “best” searches kept theirs.
What might be going on
We want to be careful to separate what we measured (the numbers above) from interpretation. A few plausible explanations, none of which we can confirm:
A reaction to the AEO/GEO gold rush. Over the past year, a large share of the marketing industry has pivoted to “Answer Engine Optimization” and “Generative Engine Optimization” — deliberately engineering content to get cited and recommended inside AI answers. Local “best X in [city]” queries are exactly where that optimization pressure is heaviest, because they’re the queries that directly produce customers. If Google’s confidence in the quality of AI answers for those queries eroded — or it detected the manipulation pressure — pulling back is a rational response.
A permanent re-scoping. Google may have simply decided that local discovery queries are better served by its traditional local results — the map pack, reviews, and business profiles — than by an AI paragraph. If so, this is the new normal.
A dip before something bigger. Google has a long history of turning features down before a major update, then relaunching them in a new form. This could be the quiet before an AI Mode expansion or a retooled local AI experience.
Anyone confidently telling you which one it is, is guessing. What we know is what we measured: the goalposts moved, for everyone, at the same time.
And this isn’t unprecedented — it’s the pattern. SEO veterans will remember Panda, Penguin, and the Helpful Content Update: overnight algorithm shifts that reshaped visibility across entire industries. AI Overviews have already had their own swings since launching. Volatility isn’t a bug in Google; it’s a recurring feature.
Why this is an argument for keeping score properly
Here’s the practical takeaway. Imagine you’re a business owner who noticed your AI-driven leads or AI visibility slipping this month. Without data like the above, the natural conclusion is we did something wrong — the site broke, the content stopped working, the agency dropped the ball. You might spend weeks and real money “fixing” a problem you don’t have.
The reality: your BOFU AI visibility didn’t dip because of anything you did. Google moved the goalposts — for every business, in every market, in the same 72 hours.
That distinction is only visible with the right measurement in place. Ad hoc spot-checks and screenshots can’t catch it: AI Overviews vary day to day, prompt to prompt, and a search that shows no AI answer on Tuesday may show one on Wednesday. One person Googling a handful of phrases and eyeballing the results is random sampling — it produces anecdotes, not evidence. Catching a systemic shift takes the same prompts, checked every day, across enough businesses and industries to separate signal from noise.
You need the right scoreboard before you can accurately keep score.
What to do about it
Don’t panic, and don’t tear up your strategy based on three weeks of data. Keep your local fundamentals strong — your business profile, reviews, and genuinely useful local content — because whatever surface Google puts in front of searchers next, those inputs feed it. And if AI answers are sending fewer people through to your site, it matters more than ever to know which businesses are actually landing on it. If AI Overviews stay gone from local queries, classic local SEO gets more valuable, not less. And if they come back in a new form, the businesses with clean measurement in place will be the first to know.
We’ll keep tracking this daily and will publish an update if the trend reverses — or if it turns out to be the leading edge of something bigger. You can find that update, and the rest of our original research, in our articles.
Are your “best in [your city]” searches still showing an AI answer?
We track the exact prompts your buyers use — every day, across Google AI Overviews, ChatGPT, Perplexity and Gemini — so a shift like this one shows up as a dated line on a chart instead of an unexplained dip in your pipeline.
Book a 20-minute walkthrough See how AI search tracking works →
Methodology
We tracked 323 search prompts across ten business accounts (anonymized) in marketing services, retail, restaurants, real estate, HR services, logistics, and home services, covering multiple U.S. markets. Each prompt was checked automatically once per day against Google (U.S.), June 22 – August 20, 2026, recording whether an AI Overview appeared (about 15,600 individual checks). Prompts were grouped by wording: “best/top” + location (137 prompts), “best/top” without a location (97), and location without “best/top” (89). The comparison chart uses only prompts tracked continuously since late June (128 / 66 / 51) so that prompts added to tracking mid-period don’t distort the trend. Percentages are daily: prompts showing an AI Overview ÷ prompts checked.