Marcel Digital review

Adequate — 3 of 5 on this rubric’s scale, from Poor (1) to Excellent (5). This dimension carries 25% of the total score.

Benchmark — across the 16 other agencies evaluated in this discipline, this dimension runs Strong 9 · Adequate 6 · Weak 1. The typical agency here scores Strong, and 9 of them score higher than this one.

What this dimension measures: What the firm actually does, described concretely enough to evaluate: content restructuring for extractability, entity and knowledge-graph work, schema, source and citation-footprint building, presence on the corpora assistants actually draw from, llms.txt and crawler access. Be sceptical of anything indistinguishable from ordinary SEO sold at a premium under a new name: if the described method is only 'good content and schema', say so plainly in the verdict rather than crediting the label.

Scores high — Named, specific techniques tied to how retrieval works score 4-5.

Scores low — 'We optimise for AI' with no described mechanism scores 1-2.

What we found — The AEO service page publishes a five-phase method: AEO readiness assessment, question and topic mapping, answer-first content optimization, schema and structured data implementation, and monitoring of AI answer surfaces. Schema types are named specifically (FAQ, HowTo, Organization, Product, Service) and entity and fact-structure work is described. A separate ChatGPT service page adds conversational keyword research and a foundational audit of where the platform sources information. What is absent is everything specific to how assistant retrieval actually works: no llms.txt, no discussion of crawler access for GPTBot or OAI-SearchBot, no off-site citation-footprint building, and no work on the third-party corpora assistants disproportionately draw from (review sites, Wikipedia, forums, industry directories). Schema plus entity clarity plus well-structured content is competent SEO practice; it is not, on the published description, a method that could only have been designed for answer engines. Capped at Adequate on that basis and because the method is self-described only. That is between the two bands, which is why it scored Adequate.

On the record — “AEO is a published service line with a five-phase method: AEO readiness assessment, question and topic mapping, answer-first content optimization, schema and structured data implementation, and monitoring of AI answer surfaces. Schema types are named as FAQ, HowTo, Organization, Product and Service.” marceldigital.com ↗

On the record — “The ChatGPT optimization page describes a five-step process and states they run real ChatGPT queries and track specific prompt performance, but names no tool, prompt set size or sampling frequency, and does not mention llms.txt, crawler access or third-party corpora.” marceldigital.com ↗

Evidence: vendor stated — the agency’s own claim, recorded as theirs rather than ours.

Adequate — 3 of 5 on this rubric’s scale, from Poor (1) to Excellent (5). This dimension carries 20% of the total score.

Benchmark — across the 16 other agencies evaluated in this discipline, this dimension runs Excellent 1 · Strong 6 · Adequate 8 · Weak 1. The typical agency here scores Adequate, and 7 of them score higher than this one.

What this dimension measures: How the firm proposes to prove anything, in a setting where outputs are non-deterministic and vary by user, session and region. Look for a described tracking approach with its limits acknowledged: prompt sets, sampling frequency, variance handling, and how a citation is attributed.

Scores high — A stated method that names non-determinism as a limit scores 4-5.

Scores low — Vague 'we track your AI visibility' scores 2-3. A firm presenting single-prompt screenshots as measurement scores 1-2 — one answer on one day is an anecdote, and treating it as a metric is the characteristic error of this category.

What we found — The service pages promise monitoring of AI Overviews, AI Mode, ChatGPT responses, Perplexity citations and Copilot answers, and the ChatGPT page states they run real ChatGPT queries and track specific prompt performance alongside referral traffic and leads. No prompt set size, sampling frequency, tracking tool or attribution rule is published. Credit is due for their own measurement article (16 Mar 2026), which is unusually candid about attribution limits and names real ones first-hand: that models synthesize information without driving a click so influence can exist without a session, that standard channel groupings rarely isolate AI sources cleanly, and that analytics only sees what happens after a visit begins. That is a genuine limit stated plainly rather than papered over. But the limit they name is attribution, not variance: nowhere on the site or in that article do they acknowledge that AI answers are non-deterministic and differ by user, session and region, which is the limit that decides whether any AI-visibility number means anything. No single-prompt screenshots are presented as proof either, so this sits above the category's characteristic error and below a described, repeatable methodology. That is between the two bands, which is why it scored Adequate.

On the record — “Their published measurement article (16 Mar 2026) names real attribution limits first-hand: models synthesize information without driving a click so influence can exist without a session, and standard channel groupings rarely isolate AI sources cleanly. It does not mention non-determinism or answer variance.” marceldigital.com ↗

On the record — “The ChatGPT optimization page describes a five-step process and states they run real ChatGPT queries and track specific prompt performance, but names no tool, prompt set size or sampling frequency, and does not mention llms.txt, crawler access or third-party corpora.” marceldigital.com ↗

attribution — working out which marketing touch actually caused a sale. Good practice names its model and its blind spots; bad practice quotes each ad platform’s self-graded numbers, which overlap and overclaim.

Evidence: partly checkable — corroborated in part against the sources below; the remainder rests on the agency’s own account.

Strong — 4 of 5 on this rubric’s scale, from Poor (1) to Excellent (5). This dimension carries 20% of the total score.

Benchmark — across the 16 other agencies evaluated in this discipline, this dimension runs Strong 4 · Adequate 8 · Weak 4. The typical agency here scores Adequate, and none scores higher than this one.

What this dimension measures: Weighted unusually high because this category's defining risk is overclaim, not opacity. No vendor controls what an assistant says. Also flag results claims presented without the prompt set or timeframe that produced them, and any implication that the firm has privileged access to a model's ranking.

Scores high — A firm that states plainly what it cannot guarantee scores 4-5.

Scores low — Guaranteed rankings, guaranteed citations, guaranteed 'first result in ChatGPT', or claimed partnerships with model providers score 1-2 and must be named in red_flags.

What we found — Across four service pages read first-hand (AEO, ChatGPT optimization, LLM ad services, monthly SEO) there is no guaranteed citation, no guaranteed ranking, no 'first result in ChatGPT' language, and no claimed partnership or privileged access with OpenAI, Anthropic, Perplexity or any model provider. The only partner badges shown are Google Premier Partner and Microsoft Advertising Partner 2026, which are established ad-platform programs, not claims about model ranking. Timelines are hedged rather than promised: early movement within a few months, depending on technical readiness, content quality and competition. The LLM ad services page is the strongest signal here, stating outright that the work is preparation for when LLM ad formats become widely available and that official LLM ad networks have not launched, rather than selling access that does not exist. Held at Strong rather than Excellent for two reasons: they never state explicitly that no vendor controls what an assistant says, and the case studies present large percentage gains (345% conversion increase, 861% learning-centre traffic) with no timeframe or baseline attached. That is the high band above, which is why it scored Strong.

On the record — “No guaranteed citation, guaranteed ranking or claimed model-provider partnership appears on any of the four service pages read. The only partner badges are Google Premier Partner and Microsoft Advertising Partner 2026, both ad-platform programs unrelated to model ranking.” marceldigital.com ↗

On the record — “The LLM advertising page sells preparation rather than access, stating the work applies 'when LLM ad formats become widely available' and that there is work to do 'even before official LLM ad networks launch'.” marceldigital.com ↗

Evidence: verified — checked against a named source you can open; the links under Sources below are where to check it yourself.

Adequate — 3 of 5 on this rubric’s scale, from Poor (1) to Excellent (5). This dimension carries 15% of the total score.

Benchmark — across the 16 other agencies evaluated in this discipline, this dimension runs Strong 10 · Adequate 6. The typical agency here scores Strong, and 10 of them score higher than this one.

What this dimension measures: Attributable client work. Be lenient on volume and strict on attribution: this category is young enough that a short roster is expected, but a long roster of unnamed logos is not evidence. Where a firm's OWN brand is demonstrably visible in AI answers, note it as a real if partial signal — it is at least a claim about themselves they cannot make privately.

Scores low — Named clients with described engagements score 4-5; anonymised or absent scores 2-3.

What we found — The general roster is real and named: SEKO Logistics, Mozilla, Johns Hopkins, Medius, Montcalm Community College, Generali USA, PerkSpot, Absolute Software, Arbour Group. But none of the twelve published case studies is about AI search, AEO, GEO or AI Overviews, and the AEO and ChatGPT service pages name no AEO client at all. For this category the attributable work is therefore effectively absent from their own site. The partial offsetting evidence is independent: their verified Clutch profile carries a client review stating the engagement produced increased AI search visibility and AI/LLM traffic, and another recording 'Capture AI Overview visibility' and 'Grow AI/LLM referral traffic' as objectives tracked ahead of plan. Those clients are described by role and sector rather than named. Their own AEO pages also surface readily in AI-assisted search, a partial self-referential signal the rubric asks to note. A strong general portfolio plus one unnamed but independently verified AI-visibility engagement lands at Adequate.

On the record — “Verified Clutch base of 4.8 out of 5 across 20 reviews, including one client reporting increased AI search visibility and AI/LLM traffic and another recording AI Overview visibility tracked ahead of plan.” clutch.co ↗

On the record — “None of the twelve published case studies concerns AI search, AEO, GEO or AI Overviews, and none of them states a timeframe for the results claimed.” marceldigital.com ↗

attribution — working out which marketing touch actually caused a sale. Good practice names its model and its blind spots; bad practice quotes each ad platform’s self-graded numbers, which overlap and overclaim.

Evidence: partly checkable — corroborated in part against the sources below; the remainder rests on the agency’s own account.

Adequate — 3 of 5 on this rubric’s scale, from Poor (1) to Excellent (5). This dimension carries 10% of the total score.

Benchmark — across the 16 other agencies evaluated in this discipline, this dimension runs Excellent 3 · Strong 4 · Adequate 2 · Weak 7. The typical agency here scores Adequate, and 7 of them score higher than this one.

What this dimension measures: Note contract length: long lock-ins are hard to justify in a category this young, where nobody can evidence a twelve-month outcome yet, and a firm offering short or month-to-month terms should be credited for it.

Scores low — Published prices, ranges or defined packages score 4-5; ranges on request score 2-3; bespoke-only scores 1-2.

What we found — No price, package or range is published anywhere on their own site, including the AEO page and the monthly SEO page, which states only that ongoing SEO is billed monthly. Recurring monthly billing is stated but no contract length or minimum term is disclosed either way, so a buyer cannot tell whether a long lock-in applies. The offsetting facts come from the independent Clutch profile, which publishes a $5,000+ minimum project size and a $200-$300 hourly rate. Scope, by contrast, is genuinely defined: the five AEO phases and the five ChatGPT steps are published deliverables, not a menu. Bespoke-only pricing on their own site, plus a defined published scope and an independently published rate range, sits at Adequate.

On the record — “No pricing, retainer range or contract term is published on their own site; the monthly SEO page states only that ongoing SEO is billed monthly. Clutch publishes a $5,000+ minimum project size and a $200-$300 hourly rate, with company size 10-49.” clutch.co ↗

deliverability — deliverability: whether cold email actually lands in the inbox rather than spam. The tell is infrastructure talk — warmed sending domains kept separate from your main domain — because a burned domain outlasts the engagement.

Evidence: partly checkable — corroborated in part against the sources below; the remainder rests on the agency’s own account.

Adequate — 3 of 5 on this rubric’s scale, from Poor (1) to Excellent (5). This dimension carries 5% of the total score.

Benchmark — across the 16 other agencies evaluated in this discipline, this dimension runs Strong 7 · Adequate 9. The typical agency here scores Adequate, and 7 of them score higher than this one.

What this dimension measures: Whether the firm acknowledges that the surfaces move, and describes how it adapts: monitoring for model and interface changes, revisiting method, published research or original testing.

Scores high — Evidence of ongoing testing and updated method scores 4-5.

Scores low — A static playbook presented as settled practice scores 2, since nothing in this category has been stable long enough to be settled.

What we found — Published output on this surface is dated and current: an AEO explainer from 23 Apr 2025, a measurement article from 16 Mar 2026, a 2026 predictions piece, and a gated 'Generative Search Playbook for B2B Brands'. The 2026 measurement piece is materially different in argument from the 2025 explainer, which is direct evidence the position is being revisited as the surfaces move rather than restated. Against that, none of it contains original testing, controlled experiments or proprietary data; it is interpretation of a changing landscape, not measurement of it. No published account of monitoring for model or interface changes on client accounts either. Recent and updated, but not evidenced testing. That is between the two bands, which is why it scored Adequate.

B2B — business-to-business: selling to companies rather than consumers — longer deals, more decision-makers.

Evidence: verified — checked against a named source you can open; the links under Sources below are where to check it yourself.

Strong — 4 of 5 on this rubric’s scale, from Poor (1) to Excellent (5). This dimension carries 5% of the total score.

Benchmark — across the 16 other agencies evaluated in this discipline, this dimension runs Excellent 2 · Strong 5 · Adequate 6 · Weak 3. The typical agency here scores Adequate, and 2 of them score higher than this one.

What this dimension measures: Independent, verified reviews. Low weight and low expectations: most firms here are under two years old, so an absent review base is a fact about the category's age rather than a finding about the vendor.

Scores low — A substantial verified base scores 4-5; a handful scores 2-3; none located scores 2.

What we found — Clutch profile read first-hand: 4.8 out of 5 across 20 reviews on a platform that requires human-led verification of reviewers. For a firm operating since 2003 this is a substantial base by the standards of this dimension. Held at Strong rather than Excellent because the reviews overwhelmingly concern SEO, paid media and web development; only one review directly addresses AI search outcomes, so the review base corroborates the firm, not specifically the AEO practice. Glassdoor ratings exist but are employee reviews and were not treated as client evidence.

On the record — “Verified Clutch base of 4.8 out of 5 across 20 reviews, including one client reporting increased AI search visibility and AI/LLM traffic and another recording AI Overview visibility tracked ahead of plan.” clutch.co ↗

Evidence: verified — checked against a named source you can open; the links under Sources below are where to check it yourself.