Seven Letter recently launched Engine, a service that audits how AI platforms like ChatGPT, Google AI Overviews, and Claude portray brands and executives. By combining quantitative reputation research with communications strategies like earned media and content development, the firm aims to influence AI-generated answers. However, the service lacks direct model control, meaning it relies on external information to sway AI outputs. Clients must distinguish between factual corrections and reputation management, as the effectiveness of these interventions remains difficult to attribute solely to the firm's efforts without rigorous, transparent benchmark testing and clear disclosure of advocacy.
Seven Letter launched Engine on October 28, 2026, offering brands, executives and issue advocates a way to assess how AI services describe them and develop a strategy to influence those answers. The communications firm calls it an Answer Intelligence service, combining prompt-based assessments with recommendations involving earned media, content and public affairs, according to its launch announcement.
The two parts are worth keeping separate. An audit can document what an AI service says under specified conditions. A communications campaign can try to improve the information available about a client. Connecting that campaign to a reliable change in AI answers requires additional evidence.
Engine’s disclosed offering is therefore best understood as reputation research paired with communications strategy. Seven Letter’s broader promise of influence should be evaluated separately from its ability to collect and analyze answers.
What Engine Actually Measures
Engine reports use custom prompts across ChatGPT, Google AI Overviews, Google AI Mode, DeepSeek, Perplexity, Copilot and Claude. Seven Letter says the reports include quantitative analysis of brand visibility, mentions and sentiment. That detail appears in the FAQ accompanying its announcement distributed through PR Newswire.
That list names AI products and search experiences, not seven fixed model versions. For buyers, the distinction is consequential: a report needs to identify the service and conditions behind each result to make later comparisons useful.
The disclosed measurements address related but different questions:
Measurement
What it can help assess
What requires separate judgment
Visibility
Whether a client appears in answers to selected prompts
Whether those prompts reflect real audience needs
Mentions
How frequently the client is named within the tested answers
Whether the mentions are relevant or influential
Sentiment
How positively or negatively an answer portrays the client
Whether the client’s communications caused the change
The last column is an analytical distinction, not a description of additional features Seven Letter has promised.
Custom prompts also make the assessment dependent on question selection. Asking an assistant to summarize a company and asking it to recommend suppliers are different reputation tests. A useful audit should separate those tasks instead of treating every mention as equivalent.
How the Audit Becomes a Communications Plan
Seven Letter’s Engine service page describes a benchmark assessment followed by a strategic action plan. Recommendations can include earned media, content and public affairs, with goals including increased visibility, factual corrections, responses to negative themes and countering misinformation.
Those channels provide several possible routes to influence:
Earned media: Seek coverage that gives outside publishers additional information about a client.
Content: Improve the client’s own published explanations, evidence and descriptions.
Public affairs: Address how an organization or issue is presented in policy discussions.
These are explanations of how the named channels could work, not a published list of Engine campaign deliverables.
The firm’s broader AI services page also advertises content strategy intended to influence generative and agentic engine results. Engine fits into that communications business: identify a representation problem, recommend changes to the surrounding information, then monitor what happens.
None of the disclosed steps amounts to direct model control. The materials describe no access to model weights, authority over an AI provider’s answer policies or guaranteed placement in responses. Publishing better information creates material an answer engine might encounter; it cannot ensure that a particular service will retrieve, trust or cite it.
I’d assess the audit and the influence campaign as separate purchases. The first can produce a useful record even when the second has uncertain results.
Why the Benchmark Design Matters
Seven Letter CEO Erik Smith says Engine gives clients “greater control over sentiment, issues and reputation,” according to the company announcement. That promise makes measurement design central to evaluating the service.
The public materials identify platforms and broad metrics, but leave important details unspecified:
Pricing and the scope of a standard engagement.
Exact prompts, prompt-selection rules and coverage of different audience needs.
Model versions, access methods and answer-generation conditions.
Repeat counts, metric definitions, sentiment scoring and weighting.
Monitoring frequency and the schedule for re-audits.
Independent validation that recommended actions change AI outputs.
These limits apply to the public descriptions reviewed here. The evidence consists of Seven Letter’s own materials, with no independent product testing supplied.
A simple hypothetical shows why the details matter. Suppose an audit uses ten fixed prompts, and a brand begins appearing in answers to two additional prompts. Its prompt-level mention rate rises by 20 percentage points. That could be useful, but the result says little about overall audience exposure unless those ten prompts represent meaningful user questions.
Changing the prompt set between audits would introduce another problem: an apparent improvement could reflect easier questions.
Even a consistent prompt set leaves attribution unresolved. A changed answer might follow a client’s new publication, unrelated news coverage, an AI product update or changes in competing information. A before-and-after comparison alone cannot distinguish those explanations.
A stronger evaluation would preserve a stable benchmark, record the exact answers and test a separate set of relevant questions that was not used to guide the campaign. It would also distinguish factual accuracy from favorable sentiment. An answer that accurately describes a documented controversy may score negatively while serving the reader well.
My preference would be to put factual errors, omissions and source quality ahead of a single positive-sentiment score. Those categories give a communications team something concrete to fix without assuming that criticism is a measurement failure.
Where Optimization Raises Disclosure Concerns
Engine’s stated goals include correcting inaccuracies and addressing negative themes. Those objectives can overlap, but they are not interchangeable. Correcting an executive’s outdated job title is straightforward. Making an accurate account of a policy dispute sound more favorable involves persuasion.
That distinction is especially important because Seven Letter explicitly markets Engine to issue advocates as well as companies and executives. Its service description places public affairs alongside content and earned media as potential tactics.
The ethical concern follows from that approach: communications work intended to influence AI answers should remain assessable as advocacy. If sponsored material presents itself as independent evidence, an answer engine could carry its framing into a response without making the underlying commercial interest clear. This is a risk analysis, not an allegation that Seven Letter uses that tactic.
Clear authorship, sponsorship disclosure and links to supporting records would help readers evaluate the underlying material. They would also help distinguish a genuine factual correction from an effort to bury valid criticism.
In my view, the useful boundary is whether a campaign improves access to checkable information or obscures who is making a claim. Seeking inclusion in an AI answer can be legitimate communications work. Concealing the sponsor behind supposedly independent support creates a different problem, even if it improves a visibility metric.
What Buyers Should Request
A sample report would be more useful than a broad promise of influence. Prospective clients should ask to see how Engine connects a prompt, its recorded answer, the assigned score and the recommended action.
For ongoing monitoring, the contract should specify the tested products, collection dates, repeat schedule and treatment of product changes. Buyers should also ask whether accuracy and sentiment are scored separately, and what evidence would count as a successful intervention.
Seven Letter says progress can be tracked through monitoring and periodic re-audits. That creates a route to evaluating the work, but the evaluation needs an agreed definition of progress. A corrected false statement, a new mention and a more flattering summary are three different outcomes. Treating them separately would make Engine’s reports more useful and its influence claims easier to assess.
Frequently Asked Questions
3 questions
1
What Is Seven Letter’s Engine?
Engine is an audit and communications-strategy service that assesses how AI products portray brands, executives and issues, then recommends actions to influence that portrayal.
2
Can Engine Directly Change an AI Model?
The disclosed service provides no direct model control. Its recommendations focus on earned media, content and public affairs, followed by monitoring.
3
Which AI Services Does Engine Assess?
Seven Letter names ChatGPT, Google AI Overviews, Google AI Mode, DeepSeek, Perplexity, Copilot and Claude. Exact model versions and testing conditions are not specified in the public materials.
