Channel strategy · A field note

GEO is the new SEO.

AI is rewriting category leadership. Buyers don't open Google's blue-link page anymore. They ask Claude, ChatGPT, Gemini, Perplexity, or read Google AI Overviews. If your brand isn't named in those answers, you don't exist in the new top-of-funnel. This is what it costs to be invisible, and how we measure it.

5 AI surfaces
~360 prompts per benchmark
4 buyer-journey stages
~24 hrs to a finished brief

Six questions your CMO is already trying to answer.

AI search isn't a feature, it's a substrate. The brand that owns the first paragraph in Google AI Overviews owns the awareness-stage slot, exactly the way SEO worked in 2010, but with a different ranking function. The AI Search Diagnostic answers the six questions any CMO is already asking, in one overnight run, with every number traceable to a raw response in the appendix.

01
Am I showing up at all?
The threshold question. % of AI answers across all five surfaces that mention your brand. Below the line, nothing else matters.
02
Where do I show up more than less?
Per-surface and per-buyer-journey-stage share. Visible in product research, invisible at awareness? That's a CMO-allocation tell.
03
Am I showing up positively, in the light I want?
Sentiment index per brand. Visibility without favorable framing erodes consideration. Worse than invisible if the framing is wrong.
04
When someone asks something negative, how am I framed?
Adversarial-prompt sentiment: what AI says when the question is about lawsuits, complaints, downsides. A 60+ point drop is a brand-fragility risk.
05
How do I show up in AI relative to organic search?
AI Overview share vs classical top-5 organic. AI is broader by construction, so some brands ride that breadth and others don't. The gap is allocation guidance.
06
Where can I attack my competition's weak content?
Source ownership audit. Typically ~95% of AI citations come from non-owned sources: affiliate, UGC, retailer, publisher. Each non-owned slot is a content lever.

deterministic Every answer is computed from the captured run; no model in the scoring loop. Raw CSVs ship in the appendix of every brief.

Five surfaces, one prompt set, your peer list.

For every benchmark we run the same ~360 prompts across all five AI surfaces in parallel, score every answer with a deterministic rubric, and roll up into a 5-KPI scorecard and an 8-page brief. The whole pipeline is reproducible: same inputs, same outputs.

Layer Signal Source Tool
01 AI surfaces Claude (Sonnet 4.5 + web_search) · ChatGPT (GPT-4.1 + web_search_preview) · Gemini (2.5 Flash + Google Search grounding) · Perplexity (sonar-pro) · Google AI Overviews (SerpAPI) Native APIs
02 Prompt set ~360 prompts, machine-generated and human-audited, spanning 4 buyer-journey stages: awareness · consideration · evaluation · purchase Haiku promptgen
03 Peer set 5–6 brands you choose as the competitive set. Optionally, the diagnostic also flags surprise competitors: brands AI named that you didn't list. Customer-supplied
04 KPI scoring Deterministic rubric over captured responses: mention rate, source diversity, citation frequency, sentiment, per-surface share Pure-function lib/scoring.ts
05 Output 8-page CMO brief + raw CSVs (responses, citations, sentiment, AI Overview share, discovered brands) EJS + d3 SVG
Claude · ChatGPT · Gemini · Perplexity · Google AI Overviews · ~360 prompts · overnight runtime · 8-page brief

Three charts that change how a CMO allocates next quarter.

We ran the diagnostic on Scotts Miracle-Gro vs a six-brand peer set (Pennington · Spectracide · Milorganite · Espoma · Jonathan Green · Sunday). Below: three of the actual charts from the live brief, with the data live in this page. Hover any bar for the exact number; tap the explainer to decode it.

01
Where in the buyer journey does AI surface Scotts?
% of AI answers naming Scotts vs. peer average, by buyer-journey stage. Source: prompt_matrix.csv · 90 prompts × 4 surfaces × 7 brands.
Share of AI answers naming each brand, by stage × surface
prompt_matrix.csv · n = 360
Claude
ChatGPT
Gemini
AI Overview
Awareness
0%
0%
0%
0%
Problem identification
0%
0%
0%
0%
Safety / health
0%
0%
17%
20%
Product research
25%
33%
33%
27%
Comparison
0%
0%
0%
0%
Post-purchase
0%
20%
0%
20%
Claude
ChatGPT
Gemini
AI Overview
Awareness
0%
0%
0%
0%
Problem identification
0%
0%
0%
0%
Safety / health
0%
17%
17%
0%
Product research
17%
11%
17%
18%
Comparison
0%
0%
0%
0%
Post-purchase
0%
0%
0%
0%
Claude
ChatGPT
Gemini
AI Overview
Awareness
0%
0%
0%
0%
Problem identification
0%
0%
0%
0%
Safety / health
0%
0%
17%
20%
Product research
17%
0%
17%
18%
Comparison
0%
0%
0%
0%
Post-purchase
0%
0%
0%
0%
Claude
ChatGPT
Gemini
AI Overview
Awareness
0%
0%
0%
0%
Problem identification
0%
0%
0%
0%
Safety / health
0%
0%
0%
40%
Product research
8%
11%
8%
9%
Comparison
0%
0%
0%
0%
Post-purchase
0%
0%
0%
0%
StageScottsPeer avgGap
Awareness 0% 0% +0 pts
Problem identification 0% 0% +0 pts
Safety / health 9% 4% +4 pts
Product research 36% 11% +25 pts
Comparison 0% 0% +0 pts
Post-purchase 10% 0% +10 pts
Decode

What you're looking at: mention rate for Scotts (and three peers) at each Consumer Decision Journey stage, broken out by AI surface. The bottom table shows the stage-by-stage gap between Scotts and the peer average.

Why it matters: Scotts dominates product research (+25 pts vs peers) but is invisible at awareness and comparison. Marketing dollars going to awareness ad spend are landing in a stage AI doesn't even surface brands in. The fix is upstream content, not more ads.

02
Where is AI getting what it says about Scotts?
Share of cited sources by ownership class, with the actual top domains AI is pulling from. Source: citations.csv, geo_targets.csv · 2,640 URLs across 4 surfaces.
Share of AI-cited sources, by ownership class
citations.csv · brand-level breakdowns
Brand-owned 4% scottsmiraclegro.com, scotts.com, scotts.pissedconsumer.com
Brand-owned 9% espoma.com, espomapro.com
Affiliate / review 16% bobvila.com, garden.org, whygoodnature.com
UGC 23% youtube.com, facebook.com, reddit.com
Publisher 6% topclassactions.com, classaction.org, mensjournal.com
Retailer 15% homedepot.com, walmart.com, domyown.com
Other 31% marijuanapassion.com, deneweths.com, meadowsfarms.com
Seed: 308 citations from responses where Espoma was named (across all 90 prompts × 4 AI surfaces)
Brand-owned 11% milorganite.com, milorganite.org
Affiliate / review 12% lawnphix.com, lawnlove.com, bobvila.com
UGC 28% youtube.com, reddit.com, thelawnforum.com
Publisher 11% jimmylewismows.com, beyondpesticides.org, mensjournal.com
Retailer 11% homedepot.com, walmart.com, growcycle.com
Other 27% radicalmoderate.online, growingagreenerworld.com, alphaoneinc.com
Seed: 310 citations from responses where Milorganite was named (across all 90 prompts × 4 AI surfaces)
Brand-owned 0% company-owned domains, PDPs
Affiliate / review 16% bobvila.com, todayshomeowner.com, eartheasy.com
UGC 16% youtube.com, reddit.com, facebook.com
Publisher 18% getsunday.com, scottsmiraclegro.com, trugreen.com
Retailer 7% homedepot.com, samsclub.com, amazon.com
Other 43% betterhelp.com, wikipedia.org, reachlink.com
Seed: 227 citations from responses where Sunday was named (across all 90 prompts × 4 AI surfaces)
Per-class quote panels are scraped only for the target brand (Scotts) in this run. Click the Scotts tab above to see top quotes by ownership class.
Click a segment or legend row above to see the top quotes AI is pulling from that source class.
Top quotes · Affiliate / review
Highest-frequency excerpts AI surfaces pulled from this source class. Sampled from citations.csv, joined with the scraped quote corpus.
  • lawnchick.com · Claude · about Scotts
    "In today’s article I’ll compare Scotts vs Pennington lawn care products to help you determine which brand might be best for you in terms of grass seed, fertilizer, and the best weed and feed products."
    https://lawnchick.com/scotts-vs-pennington/
  • greensidelawncare.com · Claude · about Scotts
    "We decided to test: * Scott’s EZ Seed * Pennington One Step Seed * Jaguar Tall Fescue Seed * 5 Way Blend Fescue Seed All products were purchased at a local store – we did not contact Scotts or Pennington."
    https://greensidelawncare.com/scotts-vs-pennington-review/
  • freeplants.com · Claude · about Scotts
    "The decision to purchase Scotts or Pennington’s grass seeds generally depends on what is it that you need and how knowledgeable and how much experience you have in lawn maintenance."
    https://www.freeplants.com/scotts-vs-pennington-grass-seed/
  • bobvila.com · Claude · about Scotts
    "Scotts Turf Builder is one of the most popular and economical fertilizer and weed control combination products available."
    https://www.bobvila.com/reviews/best-weed-and-feed-2026/
  • todayshomeowner.com · Claude · about Scotts
    "Scotts My Lawn app : The My Lawn app from Scotts allows you to enter information about your grass type, lawn size, and your location, and then you will get recommendations on what you need and when to apply it."
    https://todayshomeowner.com/lawn-garden/guides/trugeen-vs-scotts/
Source · top_quotes.csv (web-scraped from citation URLs) Showing top 5 of 93 cited URLs
Top quotes · UGC
Highest-frequency excerpts AI surfaces pulled from this source class. Sampled from citations.csv, joined with the scraped quote corpus.
Source · top_quotes.csv (web-scraped from citation URLs) Showing top 3 of 157 cited URLs
Top quotes · Publisher
Highest-frequency excerpts AI surfaces pulled from this source class. Sampled from citations.csv, joined with the scraped quote corpus.
Source · top_quotes.csv (web-scraped from citation URLs) Showing top 5 of 58 cited URLs
Top quotes · Retailer
Highest-frequency excerpts AI surfaces pulled from this source class. Sampled from citations.csv, joined with the scraped quote corpus.
Source · top_quotes.csv (web-scraped from citation URLs) Showing top 6 of 50 cited URLs
Top quotes · Other
Highest-frequency excerpts AI surfaces pulled from this source class. Sampled from citations.csv, joined with the scraped quote corpus.
Source · top_quotes.csv (web-scraped from citation URLs) Showing top 6 of 134 cited URLs

SEO is the 4%. Generative-engine reach is the other 96%.

SEO budgets target the slice Scotts owns directly. Generative engines pull most of their substance from affiliate review, UGC, publisher and retailer ecosystems. Each percentage point of UGC or affiliate citation share is leverage on AI visibility that classical SEO does not buy.

Source breadth · How evenly AI's sources are spread across the 6 ownership classes

Range 0–100. High = AI sees the brand across the whole web; Low = AI knows it from one corner. Narrow source bases collapse when a single publisher reframes the brand.

BrandSource breadth
Spectracide 95
Pennington 95
Milorganite 94
Espoma 93
Jonathan Green 92
Scotts 91
Sunday 81

How it's computed: normalized Shannon entropy of cited URLs across the 6 ownership classes — −Σ pi·log2(pi) / log2(6) × 100, where pi is the share of each class. 100 = perfectly even (16.7% in each class); 0 = every citation from a single class. From brand_scores.csv.

GEO priority grid · sentiment × cited

Cross-cut of the audited GEO target pages by (a) whether Scotts is on the page and (b) page-level sentiment toward whoever IS named. Click a cell to expand the top 5 URLs in that quadrant. From 212 audited pages. Sentiment via Haiku, page bodies via Bright Data.

Scotts on page · cited
Scotts NOT on page · open
Positive sentiment
Negative sentiment

Defense priority · Scotts cited with negative sentiment

Pages where AI is pulling negative framing about your brand. Highest leverage to correct — invest in alternative content that ranks for the same queries, or directly engage the page (response, correction).

  1. https://www.oasisturf.com/blog/4-big-problems-with-the-scotts-4-step-lawn-care-approach
    4× cites · Claude, AI Overview · publisher
  2. https://pennington-seed.pissedconsumer.com/review.html
    6× cites · Claude · ugc · peers (positive): Pennington, Jonathan Green
  3. https://pennington-seed.pissedconsumer.com/reviews/RT-P.html
    6× cites · Claude · ugc · peers (negative): Pennington
  4. https://www.trustpilot.com/review/getsunday.com
    2× cites · Claude · affiliate · peers (negative): Sunday

Offense priority · peers cited positively, Scotts absent

Pages where competitors get favorable framing and your brand isn't named at all. Highest leverage to plant — sponsor placement, earn an edit, or build the comparable page yourself.

  1. https://www.thelawnforum.com/threads/milorganite-what-do-you-guys-think-of-it.12336/
    4× cites · Claude, AI Overview · ugc · peers (positive): Milorganite
  2. https://www.houzz.com/discussions/5369018/espoma-and-miracle-grow-fertilizer
    2× cites · Claude, AI Overview · ugc · peers (positive): Espoma
  3. https://www.milorganite.com/blog/garden-landscape/dog-friendly-landscape-ideas
    2× cites · Claude, AI Overview · publisher · peers (positive): Milorganite
  4. https://www.indeed.com/cmp/The-Pennington-School/reviews
    3× cites · Claude · ugc · peers (positive): Pennington
  5. https://www.jonathangreen.com/product/organic-lawn-food/
    2× cites · Claude · publisher · peers (positive): Jonathan Green

Already winning · Scotts cited positively

Sources where AI describes your brand favorably. Maintain the relationship; don't over-invest where you're already winning.

  1. https://www.yourgreenpal.com/blog/pet-safe-lawn-fertilizer
    2× cites · Claude, AI Overview · affiliate · peers (positive): Milorganite · peers (negative): Espoma
  2. https://lawnphix.com/lawn-care/organic-fertilizer-for-lawn/
    2× cites · Claude, AI Overview · affiliate · peers (positive): Milorganite, Espoma
  3. https://www.bobvila.com/reviews/best-liquid-lawn-fertilizer/
    2× cites · Claude · affiliate
  4. https://lawnbyseason.com/best-lawn-fertilizer
    2× cites · Claude · affiliate · peers (positive): Pennington, Milorganite, Jonathan Green

Skip · competitor takedowns

Pages where peers are described negatively and you're not on them. Trying to plant your brand here would associate it with the takedown — leave alone.

  1. https://jimmylewismows.com/blog/no-more-milo
    8× cites · Claude, AI Overview · publisher · peers (negative): Milorganite
  2. https://www.sitejabber.com/reviews/penningtons.com
    5× cites · Claude, ChatGPT · affiliate · peers (negative): Pennington
  3. https://www.domyown.com/is-spectracide-triazicide-insect-killer-for-lawns-and-landscapes-safe-for-do…
    2× cites · Claude, AI Overview · retailer · peers (negative): Spectracide
  4. https://www.thcfarmer.com/threads/has-anyone-tried-this-product.74380/
    4× cites · Claude · ugc · peers (negative): Espoma
  5. https://www.jonathangreen.com/shop/
    3× cites · Claude · publisher · peers (negative): Jonathan Green

Full ranked list of every cited URL — class, surfaces, cite count, audit verdict — is in geo_targets.csv in the data folder.

Source · citations.csv · geo_targets.csv · 2640 URLs classified across 4 surfaces Page 05 / 08
Decode

What you're looking at: the full source-ownership ring for what AI cites about Scotts, with the actual top domains in each class. Toggle the brand switcher at the top to see how Scotts compares to peers.

Why it matters: Scotts owns just 4% of what AI says about it. The other 96% is third-party: review sites, YouTube/Reddit, retailers, publishers. Classical SEO budgets target the 4%. AI visibility is bought in the 96%. Each percentage point of UGC or affiliate share is leverage classical SEO can't buy.

03
How does each brand hold up when AI is asked something negative?
Sentiment index, ambient (default prompts) vs. adversarial (downsides / lawsuits / complaints). Source: brand_scores.csv.
Sentiment Index, by brand · ambient vs adversarial
brand_scores.csv
Adversarial Ambient 3333676710010000 20 100 Spectracide 30 100 Milorganite 42 100 Espoma 27 91 Scotts 50 88 Jonathan Green 40 75 Sunday 20 58 Pennington
BrandAmbientAdversarialResilience
Spectracide 100 20 -80
Milorganite 100 30 -70
Espoma 100 42 -58
Scotts 91 27 -64
Jonathan Green 88 50 -37
Sunday 75 40 -35
Pennington 58 20 -39
Source · brand_scores.csv · sentiment classified across all collected responses Page 06 / 08
Decode

What you're looking at: each brand's sentiment under default AI prompts (ambient) vs. when AI is primed with negative framing (adversarial: lawsuits, complaints, product downsides). The "Resilience" column is the drop.

Why it matters: Every brand drops under adversarial prompts. How far it drops is the brand-fragility tell. Scotts drops 64 pts when AI is primed negatively. Pennington starts lower but holds up better. Resilience > 60-pt drop = a fragility risk worth flagging to comms/legal.

deterministic Numbers above are pulled live from the Scotts brief generated 2026-05-08 · 360 prompts × 4 surfaces · charts re-rendered natively in this page so they're searchable, copy-pasteable, and accessible. Full 8-page brief with raw CSVs: aisearchdiagnostic.com/output/lawn-care-scotts-focus-2026-05-08 →

GEO is the new SEO

This is your jumping-off point.

Run a benchmark on your brand. Pick a peer set, pick a category, and we'll generate the same 8-page brief: same five surfaces, same scoring rubric, same auditable CSVs. ~24 hours from intake to delivered brief.