The central debate in Generative Engine Optimization (GEO) right now is deceptively simple: does AI Search care more about who you are on your own site, or who talks about you across the web?
OtterlyAI designed a controlled, real-world experiment to find out. We built a brand from scratch, zero domain authority, seven pages, and systematically tested whether off-page authority signals could drive meaningful AI Search visibility inside a competitive market, outpacing brands with far more history, content, and domain presence.
Key Takeaways (TL;DR)
- A new brand with zero domain authority, low number of pages can outperform an existing brand if mentioned on enough external sites that are authoritative to AI Search.
- 16 AI citations is all it took: Within 14 days this new brand reached 10% share of voice, landing at rank #7 out of 10+ competitors , purely through 16 brand mentions on external sites & directories.
- Google Maps not needed: There was no possibility in this experiment to create a Google Business page in London without Google’s verification. However, Google Maps was cited for other brands.
- The importance of Exact Match Blogs (EMBs): The brand only appeared in AI results when the query closely matched the article’s exact phrasing, such as “Top GEO agencies 2026 in London.” Broader prompts like “Top GEO agencies” did not trigger visibility, showing EMB content mainly ranks for the exact Search Prompt it targets.
- AI Search platforms measure authority differently: The data showed that for Copilot and AI Mode On-page SEO signals and domain authority are still very important when recommending brands.
- Directory listings still proved to be the most impactful way to affect AI visibility. However, LinkedIn posts and even Reddit thread replies were also being picked up by the AI Search Platforms. And the AI also took the info of the brand’s own website into consideration.
Why This Experiment Matters for GEO
AI Search does not rank pages. It cites sources. That fundamental shift changes the entire calculus of how a brand earns visibility , and it means that the traditional SEO playbook, which focuses almost entirely on on-page signals, may leave significant AI visibility on the table.
The GEO community lacks controlled experiments isolating off-page versus on-page contributions to AI citation frequency. Most case studies blend both. This experiment was designed to isolate the variable: a brand with intentionally minimal on-page presence, built specifically to test whether external citation signals alone could drive AI brand mentions.
Hypothesis
Off-page authority signals are more impactful than on-page signals when it comes to increasing AI Search visibility.
The specific theory tested: a website with no domain authority and only 7 pages can still outperform brands that have been operating longer , if given enough brand mentions on external websites that AI Search platforms already cite heavily.
Two interpretive notes apply throughout:
- Correlation does not imply causation.
- Results reflect a 14-day window in one market (“GEO agencies in London”) and should be treated as directional, not universally prescriptive.
Methodology:
Step 1: Define the Scope and Select a Blue Ocean Market
The experiment needed a market where outside factors , existing brand equity, high-authority competitors, algorithmic noise , could be minimized as much as possible. “GEO agencies in London” was selected as the target because:
- GEO as a category is new enough that no brand dominates AI citations
- The London qualifier adds geographic specificity that limits broader competition
- Prompt research confirmed that AI engines actively return agency lists for this query type
Selecting a blue ocean market framing was deliberate. Testing in a saturated market (e.g., “SEO agencies globally”) would have made it nearly impossible to isolate the impact of off-page signal building within a 14-day window.
Step 2: Build a Minimal Brand But Have Exact Info
A new fictional GEO agency (Agency X) was created from scratch. The site was designed with the minimum viable number of pages needed for AI platforms to understand what the brand is and does:
- Homepage (who we are, what we do)
- Services page
- About page
- Contact page
- Three supporting content pages
- The footer however, explicitly states this brand is “a GEO Agency in London.”
- No Schema markup was used in this experiment
No blog. No link profile. No domain history. The on-page footprint was intentionally kept small to ensure that any visibility gains tracked during the experiment could not be attributed to content depth or traditional SEO authority.
Step 3: Conduct Prompt Research & Configure AI Search Monitoring
Before launching any off-page activity, a full prompt research session was conducted to identify the queries AI platforms were already using to surface GEO agency lists. These prompts were then loaded into an OtterlyAI Brand Report to establish a clean baseline and enable daily tracking of brand mention frequency, share of voice, brand coverage, and average brand position.
The monitoring setup covered all major AI Search engines: ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, Microsoft Copilot, and Gemini. The 9 tracked prompts included variations on location, audience, and year , for example:
- “Top GEO agencies 2026 in London?”
- “What are the top GEO SaaS Agencies of London 2026?”
- “Best GEO Agencies in the UK?”
Step 4: Identify Top AI Citation Opportunities
The next step was identifying which external surfaces AI Search platforms actually cite when answering these types of queries. Based on OtterlyAI’s citation database, the highest-leverage content types for agency/service-category queries are:
- Blog listicles on SEO/marketing news sites
- Reddit threads discussing agency recommendations
- LinkedIn Pulse articles written by practitioners
- PR placements on industry publications (like OnSaaS)
- Directory listings on platforms like Sortlist, Clutch, Goodfirms, and DesignRush
Each of these surface types has documented citation frequency across AI platforms , Reddit and LinkedIn in particular are consistently among the top cited social domains in AI Search responses.
Step 5: Create Tradeable Listicle Assets
Rather than only asking for brand mentions, a set of original listicle pages was created , “Top GEO Tools,” “Best GEO Resources for Agencies,” and similar , that could be offered as a direct value exchange to other sites already ranking with their own listicles.
Alternatively if you have high authoritative websites in your network, you can ask them to create listicles as well.
This approach transforms a cold outreach campaign into a reciprocal trade: your brand gets a mention on their listicle, and they get a useful resource to link to or embed on theirs. It removes the friction of pure link-begging and frames the request around mutual visibility.
Step 6: Outreach , Trade Listicle Placements for Brand Mentions
Outreach was sent to owners and editors of existing GEO/SEO agency listicles offering to swap: a backlink or mention of their resource in exchange for a mention of Agency X on their own list.
The outreach targeted websites already confirmed to appear in AI citations for the target prompts , not just high-DA sites, but specifically the sources AI was already pulling from. This distinction matters enormously. A placement on a site that AI engines don’t cite has near-zero AI visibility value, regardless of its traditional SEO authority.
The outreach process is the only part that is very laboursome. For this experiment countless manual outreach messages were sent, less than 16 websites agreed for a link exchange. If I had to redo this experiment, I would use automation tools.
Pro Tip: Tools like Noble will help you automate the blog outreach process. This is an example of how that looks like:
Step 7: Engage Social and UGC Surfaces
Two categories of social engagement were executed simultaneously:
- Reddit: Replied to existing threads asking about GEO agencies, contributed genuine value first, and included the brand as one of several legitimate recommendations. New threads were not created , only genuine participation in existing discussions.
- LinkedIn Pulse: Published short practitioner articles (500–800 words) authored under the brand, covering topics aligned with the tracked prompts , e.g., “What to look for in a GEO agency in 2026.”
These placements were chosen based on citation evidence: Reddit is the single most cited social domain in AI Search (46.4% of social media citations), and LinkedIn accounts for 13% , together they represent the majority of social citation surface area for professional queries.
Tip: Learn how Social Media affects AI Visibility, read the YouTube Citation Study 2026
Step 8: Create Free Directory Listings
The brand was registered on all major free directories known to be cited by AI Search platforms:
- Sortlist
- Clutch
- Goodfirms
- DesignRush
- And more
Each listing was optimized with consistent brand description, service category tags, and London/UK location data, even though the address was not a real location. These platforms rank highly in AI citations for “best agency” and “top agency” query types because they are structured, authoritative comparison sites , exactly the format AI prefers when answering discovery queries.
Additionally there were also paid directories around. But we didn’t include those in this short experiment.
Step 9: Measure and Analyze AI Brand Visibility Over 14 Days
The OtterlyAI Brand Report ran continuously across the 14-day window, with all 9 prompts tracked daily across all major AI platforms. The key metrics tracked were:
- Brand Mentions , total times the brand appeared in AI-generated responses
- Share of Voice , the brand’s mention share relative to all detected competitors
- Brand Coverage , percentage of tracked prompts where the brand appeared in the answer
- Average Brand Position , average rank position when mentioned
Results
All AI Search Platforms:
After 14 days of off-page activity , with no changes to the brand’s website after launch , the results were as follows.
Brand Ranking (Data from OtterlyAI and Semrush):
| # | Brand | DA | Indexed pages | Keywords in Top 3 (UK) | AI Mentions | Share of Voice | Brand Coverage |
| 1 | NoGood | 41 (very good) | 7.850 | 50 | 115 | 13% | 15% |
| 2 | SEO Works | 42 (very good) | 556 | 1021 | 113 | 12% | 15% |
| 3 | Blue Array | 41(very good) | 458 | 50 | 112 | 12% | 15% |
| 4 | Pearl Lemon | 36 (very good) | 40.400 | 7 | 100 | 11% | 13% |
| 5 | Ignite SEO | 35 (very good) | 57 | 21 | 96 | 11% | 13% |
| 6 | MADX | 24 (average) | 238 | 1 | 91 | 10% | 12% |
| 7 | Agency X | 0 (poor) | 7 | 0 | 90 | 10% | 12% |
| 8 | Bird | 33 (very good) | 700 | 1 | 85 | 9% | 11% |
| 9 | Skale | 25 (good) | 300 | 0 | 69 | 8% | 9% |
| 10 | SaaStorm | 22 (average) | 122 | 0 | 22 | 2% | 3% |
Two prompts drove 74% of all brand mentions: location-specific, year-tagged queries. The non-London and non-year-tagged prompts delivered near-zero results, even with the same off-page activity in place.
Average Brand Position: 1.71
When The brand appeared in an AI response, it appeared near the top of the list , averaging position 1.71 across all prompts. Competitor benchmarks: AWISEE (1.0), Flow (1.31), Ignite SEO (1.51).
ChatGPT
ChatGTP showed significant preference for off-page authority signals, ranking the brand in the top 3 of the AI brand leaderboard.
Google AI Overviews:
The Google AI overviews showed a very similar pattern, even ranking the brand on position 2 of the leaderboard.
Google AI Mode:
Google AI Mode however rewarded brands that had strong on-page SEO signals and high domain authority. On this platform we ranked on the 8th place.
Perplexity:
Medium impact was noticed in Perplexity, however the brand still ranked in the top 5.
Copilot:
Microsoft copilot clearly works in a very different way. In this scenario the brand was not performing well. Other signals seem to prioritize other brands here.
Gemini
Gemini was the only outlier from all the AI platforms. Here we saw strange patterns in the recommendations. It hardly recommended any of the brands listed in our brand tracker.
Discussion: What This Means for How You Do GEO
Off-Page Signals Are the Primary Driver of Early AI Visibility
The most significant finding is the most direct one: a 7-page site with no domain authority entered a competitive AI query category at rank 7 out of 10+ tracked brands within 14 days. The only variable was off-page citation activity , directory listings, social engagement, and listicle placements on sites AI already cites. On-page content was intentionally minimal and unchanged after launch.
This does not mean on-page content is irrelevant. It means that for early AI visibility in defined market categories, the fastest lever is being mentioned on the external sources AI Search platforms trust , not building more pages on your own site.
Some AI Search Engines Still Reward Classic SEO signals
This experiment also shows that not all AI Search engines rank brands in the same way.
- Google AI Mode and Microsoft Copilot clearly show a preference for brands that have higher domain authority and more content that performs well in SEO.
- ChatGPT & Google AI Overviews clearly rewards external authority as it ranked our brand
- Perplexity showed in this experiment mixed results. It did place the brand in the top 10; but the results were not so clear like it was in ChatGPT. It seems that Perplexity looks for other signals to rank brands.
- Gemini seems to be the only AI Search platform that shows a strange behavior, hardly any of our listed brands were cited at all.
AI Search Platforms Are Citation Networks, Not Ranking Engines
Traditional SEO rewards the best-optimized page. AI Search rewards the best-cited brand. The implication is structural: the question is not “how do I rank my page?” but “which external sources does AI pull from, and how do I get mentioned there?”
OtterlyAI’s citation data confirms that directories, Reddit, and LinkedIn consistently appear among the most-cited domains for discovery-type queries. Being present and accurate on those surfaces is more directly connected to AI visibility than adding a blog post or optimizing a title tag.
Prompt Specificity Is a Targeting Mechanism
76% of the brand’s mentions came from two prompts , both of which were location-specific (“in London”) and year-tagged (“2026”). The broader or less specific prompts (“Best SEO Agencies in London,” “Top SaaS GEO Agency for Tech Startups?”) returned zero mentions despite identical off-page activity.
This suggests that AI Search behaves like a highly specific retrieval system: the more precisely a prompt matches the combination of your category, location, and recency signals, the more likely your brand surfaces. GEO strategy should map prompt types explicitly rather than chasing general keyword coverage.
This confirms the strategy of having Exact Match Blogs (EMB’s) works very well. AI Search is very much about providing the best answers as specific as possible.
The 14-Day Acceleration Window
Brand coverage was largely flat for the first week of the experiment, then accelerated sharply in the final five days. This pattern likely reflects the indexing and re-crawl cycles of AI platforms processing newly published external content , directories, LinkedIn articles, and Reddit threads don’t generate AI citations instantaneously. There is a propagation lag, and strategies that don’t account for it will underestimate the eventual impact of off-page activity.
What Off-Page GEO Doesn’t Cover
The experiment intentionally isolated off-page variables, which means it also revealed the edges of what off-page work alone can do. Zero brand mentions were generated for the non-location, non-year prompts. Broader visibility , “Top SaaS GEO Agency for Tech Startups?” or “Best SEO Agencies in London” , likely requires either stronger on-page topical authority, or a much longer citation-building timeline, or both. Off-page signals are a fast-start lever, not a complete strategy.
The GEO Off-Page Playbook: What to Actually Do
Based on what this experiment demonstrated, here is what to prioritize if your goal is faster AI brand visibility.
1. Map the citations before you build content. Identify which external sources AI Search platforms cite for your target prompts before deciding where to invest time. A placement on a non-cited site contributes nothing to AI visibility. Use an AI Search monitoring tool to trace citation sources directly.
2. Start with directories. Sortlist, Clutch, Goodfirms, and DesignRush appeared as citation-relevant surfaces in this experiment. They are free, fast to set up, and consistently referenced by AI platforms for agency and professional-services queries. If you are not listed and optimized on these platforms, fix that before anything else.
3. Engage Reddit and LinkedIn authentically. Reddit is the most cited social domain in AI Search (46.4% of all social citations). LinkedIn accounts for 13%. Both reward genuine participation , not promotional spam. Contributing real value to existing threads and publishing practitioner-level articles generates citation-eligible content that AI surfaces in discovery queries.
4. Target listicles on already-cited sites. Getting named on an existing agency listicle that AI already pulls from is more valuable than creating your own. Outreach with a value exchange , a tradeable listicle, a resource mention, a guest contribution , converts cold asks into reciprocal placements.
5. Use prompt-specific language in your external mentions. The experiment showed that location + year tagging is a signal multiplier. When your brand is mentioned in content that already contains the phrasing AI systems are retrieving , “best GEO agency in London 2026” , citation likelihood increases. Make sure directory profiles, article titles, and social posts reflect the exact language of your target prompts.
6. Be patient through the propagation lag. External citations do not generate AI visibility immediately. Plan for a 7–10 day lag between publishing and seeing measurable brand coverage movement. Don’t stop a campaign because early results are flat.
Closing Thoughts
This experiment confirms something that anyone building in the AI Search space needs to internalize: AI citation behavior looks more like a reference network than a ranking engine. Most AI Search platforms don’t reward the biggest website or the most optimized page , they reward the most cited brand across sources they already trust.
A 7-page site with zero domain authority, built in a week, reached rank 7 in its target category within 14 days of off-page activity. That is not a fluke , confirming external validation is a structural property of how AI Search retrieves and surfaces brands.
The practical implication is direct: if your GEO strategy is primarily on-page, you are optimizing for the wrong signal. The first question to ask is not “what should I write?” but “where does AI already look, and is my brand named there?”
If you want to track your own AI Search brand visibility and run experiments like this one, explore OtterlyAI’s Brand Report and AI Search Monitoring tools
👉 OtterlyAI’s GEO Guide , Updated for Crawlability, Entity Clarity & Citation Signals
