Turn Findings Into a Prioritized Work Plan

7.1 Diagnose Gaps in Access, Firm Facts, Expert Profiles, Content, Evidence, and Outside Representation 

Before adding a task, inspect the original observation. Confirm the question, answer, date, collection conditions, and identity of the firm or professional. Check the reference facts. An accurate statement of a limitation, or an omission in an unsuitable case, may require no corrective work.

Then inspect the relevant public assets. Look for a specific access problem, factual contradiction, missing explanation, unsupported claim, or outdated external description. Use the following six categories to organize the diagnosis. A work item can involve more than one category.

Identify the kind of gap

Access. Can the intended public information be retrieved under the conditions being tested? Identify the affected URL, requester, observed failure, and current access policy. OpenAI, for example, distinguishes search access through OAI-SearchBot from GPTBot permissions concerning possible model training. A restriction on one is not automatically a defect in the other. Check whether a restriction is deliberate before proposing a change. 

For Google’s AI Overviews and AI Mode, supporting-link eligibility requires an indexed page that is eligible for a Search snippet. Google specifies no additional technical requirements and does not guarantee crawling, indexing, or serving even when requirements are met. A verified access defect deserves attention; an unexplained omission does not establish that access is the cause. 

Firm facts. Does the public information consistently describe the firm’s identity, services, markets, engagement models, and limitations? Three pages giving different accounts of the same service create a concrete information problem. Identify the approved description before assigning someone to rewrite them.

Expert profiles. Are current affiliations, roles, qualifications, and relevant experience represented accurately? Check whether the information distinguishes the individual’s work from the current firm’s work. A former partner still listed as the practice leader is a verifiable defect in an active team directory.

Content. Does an appropriate asset answer the client’s actual question? A practice page might explain a methodology while omitting the engagement’s deliverables, intended clients, or scope. State the missing decision-relevant information. “We need more content” is too broad to assign.

Evidence. Can material claims be supported with information the firm may legitimately publish? Distinguish evidence that does not exist from evidence that exists but lacks publication permission. The next task might be verification, permission, or a more modest claim. It should not be the invention of a supporting case study.

Outside representation. Do relevant directories, associations, publications, and external biographies describe the firm accurately? Record the specific statement, its source, and who controls it. A correction request is a different deliverable from editing an owned page. Accurate criticism or an independent editorial choice may require no correction.

State what is known and what remains a hypothesis

Accounting example. An assistant recommends a practice for an audit service it no longer offers. The current service page is correct, but an old downloadable brochure still lists the service and is presented as the current brochure. That obsolete representation is a verified problem. Unless the answer cites or otherwise identifies it, its role in that answer remains uncertain; even a citation does not establish that it was the only contributing source.

Write two separate statements in the work item: “The brochure contains an obsolete service claim” and “Updating or retiring it may reduce future misrepresentation.” The first is a finding. The second is the expected effect of an intervention.

Leave a cause-unresolved category available. A cited competitor’s report can be useful to inspect without proving that its format caused the citation. Bing itself describes citation share as observational and identifies several influences on changing citation patterns, including model changes, user behavior, and other web changes.

Where no actionable defect is established, specify a small diagnostic task or continue observing. Link repeated symptoms to an existing item. Avoid multiplying work simply because the same issue appears in several prompt variants.

7.2 Prioritize Improvements by Client Relevance, Business Value, Confidence, Effort, and Risk

Begin with two decisions that precede ordinary ranking. First, does a potentially material misrepresentation require prompt verification or correction? Second, can the proposed action be supported and carried out under the firm’s publishing rules? A commercial opportunity does not supply missing evidence or publication rights.

For the remaining work, assess five considerations. A short rating and one sentence of explanation are usually sufficient. Use the ratings to make judgment visible; do not present them as measured probabilities of improved AI visibility.

Separate a strong case for the work from a strong prediction

Finding confidence can be high when the AI effect remains uncertain. The firm can verify that a biography states an old affiliation. Correcting that statement is justified even if the team cannot predict when assistants will adopt the new information.

For discretionary improvements, describe the benefit through the client decision. An accounting practice might prioritize a clear explanation of its transaction-support scope because prospective clients repeatedly ask what the engagement includes. That is a stronger business rationale than a general request to publish another article about a popular topic.

Do not use frequency in the prompt panel as a substitute for buyer demand. The firm selected those questions. Nor should a large competitor mention count automatically outrank a smaller problem affecting the firm’s most important service.

Google’s content guidance asks whether material serves its intended audience, contributes useful original information, demonstrates relevant expertise, and avoids factual errors. It cautions against broad volume-driven publishing and writing to a supposed preferred word count. Those principles support choosing useful assets; they do not establish the incremental GEO return from a particular revision. 

Use clear tie-break rules

After material corrections, prefer work that serves a priority client decision. Give weight to prerequisites that unlock several valuable tasks. Confirming service scope may enable a practice page, two professional profiles, and a directory correction.

Among comparably valuable items, prefer stronger evidence, lower action risk, and a deliverable that fits the scarce role’s capacity. A four-hour writing task requiring three partner hours may be harder to schedule than an eight-hour writing task requiring thirty minutes of review.

When the benefit is plausible but uncertain, select the smallest useful next commitment. Research feasibility can precede a major study. Verification can precede a case narrative. A targeted revision can precede a redesign, provided the smaller change is coherent and valuable on its own.

Record the priority as a sentence: “We will do this next because it fixes a verified misunderstanding about a priority service, enables two dependent updates, and has an available reviewer.” If the firm uses numerical scoring, keep these reasons, dependencies, and exceptions visible.

An item can be important and blocked. Preserve its strategic priority while identifying the missing evidence, permission, or reviewer. Choose another ready item for delivery rather than pretending that a high score resolves the dependency.

7.3 Separate Urgent Factual Corrections From Experiments and Longer-Term Investments

Classify the work so that each item has an appropriate decision and completion test. A verified false affiliation, a proposed change to page structure, and a new research program need different treatment. They can all contribute to GEO without sharing the same urgency or evidence requirements.

Urgent factual corrections

Potentially consequential errors warrant prompt verification by the appropriate professional. Once verified, correct the information the firm controls and identify necessary external requests. A material error does not need to recur across many answers before it matters.

Law firm example. The active team page identifies a departed partner as the current practice leader. Confirm the departure and approved replacement wording, correct the active listing, and inspect affected owned biographies and downloads. Preserve useful, permitted historical references with accurate dates and status. The owned correction can proceed while an external directory request awaits a response.

Keep the corrective scope proportionate. A verified error in one professional profile need not wait for a complete website redesign. If a potentially serious public claim remains unresolved, use the firm’s established process to review or pause that claim rather than rushing a replacement assertion into publication.

Where the firm’s own information is already accurate, assign the next controllable step: inspect a relevant external source, submit a supported correction request, or investigate the mismatch. Do not change a correct service limitation merely to make the assistant more favorable.

Bounded improvements and experiments

A bounded improvement has a finite, useful deliverable: revise a service explanation, connect a biography to relevant work, or make approved evidence easier to find. It can be justified by its client value even when its effect on AI answers remains uncertain.

Reserve the experiment label for work that tests a stated hypothesis under a defined evaluation plan. Record the proposed change, affected assets, relevant cases, observations to collect, and decision date before implementation. Chapter 17 develops the methods for assessing whether an intervention caused a change.

Consulting example. The firm hypothesizes that separating engagement deliverables from a general methodology description will improve the completeness of answers about its offer. It can revise one appropriate page, preserve the prior version, and specify the relevant monitoring cases. If other pages, prompts, or collection conditions change too, those changes belong in the record.

Do not withhold a necessary factual correction to preserve an experiment. Make the correction, record it, and reconsider the comparison. An inconclusive visibility result also does not make an accurate, useful page worthless.

Longer-term investments

Original research, publishable case evidence, substantive service resources, and sustained professional participation can require expertise and coordination over several cycles. Start with a defined client need and an evidence plan. Assign the work a sponsor, a next milestone, and capacity.

For a proposed benchmark study, the first deliverable might be a feasibility decision: the research question, permissible data, method, qualified reviewer, limitations, budget, and intended audience. Completion of that decision may justify proceeding, redesigning the study, or declining it. It does not commit the firm to publishing an unsupported report.

Protect capacity for worthwhile longer investments. A continuous queue of small edits can otherwise consume every available expert hour. Set the next commitment explicitly, while retaining the ability to respond to material corrections. The different work types draw on the same people and budget.

7.4 Decide Which Pages and Assets to Improve, Create, Consolidate, or Retire

For each information problem, identify the asset that should carry the answer. Include practice pages, professional profiles, articles, research, downloadable brochures, presentations, videos, and controlled directory listings. A corrected web page can coexist with a misleading download unless both are included in the disposition decision.

Record each affected asset’s purpose, audience, current owner, relevant evidence, and proposed action. Inspect existing links, useful traffic or referral activity, and ongoing client use before making a substantial change. Low visibility in the prompt panel alone is insufficient reason to remove useful material.

Preserve meaningful differences

Separate pages can be justified by distinct services, audiences, jurisdictions, or evidence. Similar wording alone does not settle the decision. Employment-law resources for different jurisdictions may need separate treatment because the reader’s circumstances differ.

Conversely, three consulting pages describing the same commercial due-diligence service under inconsistent names may benefit from one coherent main description. Preserve any genuinely distinct guidance. Consolidation should make the information easier to use and maintain; the resulting AI citation rate remains an outcome to observe.

Before creating a page, state the question it will answer and why an existing asset cannot do that job clearly. A new industry page needs something credible about that industry. Replacing a sector or city name in otherwise generic copy does not establish relevant experience or service coverage.

Include the URL and reference plan

A consolidation decision is incomplete until the web owner knows what happens to each existing URL. Identify the preferred destination, material being preserved, internal links to update, and checks required after publication. Keep accurate historical information where it remains useful and permitted.

Google’s migration guidance recommends mapping old URLs to appropriate new destinations and warns against sending many unrelated pages to an irrelevant home page. Where content has genuinely been consolidated, redirects to that consolidated page can be appropriate. Apply this mapping principle to the affected assets without turning every edit into a site migration.

If substantially similar pages need to remain available, ask the technical owner whether a canonical preference is appropriate. Google describes redirects and canonical annotations as signals for selecting among duplicate or very similar pages. A canonical setting does not correct contradictory copy or guarantee how every AI system will treat it. Chapter 10 covers implementation. 

Choose substantive changes

An old date does not itself make information false. A completed project may remain a useful historical example; a former role can remain part of an accurately dated career record. Assess whether the asset’s current presentation is misleading, whether its substance remains useful, and whether publication is still permitted.

Google specifically cautions against changing dates without substantive changes and against mass additions or removals intended merely to make a site appear fresh. Give each proposed refresh or retirement a content reason that can be inspected.

When evidence cannot be published, reconsider the asset. An approved account, a supported aggregate finding, or an explanation of method may serve the client better than an unusable case study. Removing a client name does not by itself settle confidentiality or permission. Keep the decision within Chapter 4’s rules.

7.5 Assign Owners, Expert Reviewers, Deliverables, Dependencies, and Acceptance Criteria 

Give each selected item a brief that another person could execute without reconstructing the discussion. Keep the main backlog readable and put supporting detail in the linked item. One accountable owner should coordinate completion; professional reviewers and publishers should have explicit responsibilities.

Assign people, not departments. “Marketing” is an incomplete commitment. Name the person responsible for delivery, the professional who can approve the claims, and the person authorized to implement the change. In a small firm, one person may fill several roles.

Use a consistent work-item brief

1. Finding and evidence. Give the item an ID. Link relevant observation records and assets, state the verified problem, and preserve the remaining hypothesis.
2. Purpose and priority. Identify the client decision, practice or market, work type, and reason for doing it now.
3. Next deliverable. Specify an inspectable output: one revised service page, a corrected profile set, a URL consolidation, a research feasibility decision, or a supported external request.
4. People and dependencies. Name the owner, required reviewer, and publisher. List missing facts, evidence, permissions, technical inputs, and prerequisite items.
5. Effort and dates. Estimate time by role, separate working time from outside waiting time, and identify the next milestone.
6. Acceptance criteria. State the factual, editorial, permission, and technical checks the deliverable must pass.
7. Expected outcome and observation. Describe the intended client benefit and the hypothesized AI effect. Link the monitoring cases and assign a follow-up date.
8. Maintenance and status. Name the ongoing asset owner, review trigger or date, and current delivery status.

Define acceptance in terms the firm can verify

For a practice page, acceptance might require approved service scope and exclusions, supported claims, an appropriate expert review, working links, and verified access under the defined checks. For a profile, require the current affiliation and approved qualifications to match the reference records. Inspect the published version as well as the draft.

For external representation, distinguish a complete correction request from the outside publisher’s response. The firm can verify what it submitted, the supporting evidence, submission date, and follow-up owner. “Directory corrected by Friday” is not a controllable delivery commitment when another organization decides what to publish.

Likewise, “the approved page is published and accessible” is different from “ChatGPT recommends the firm.” Give implementation acceptance and subsequent observation separate fields. The firm may set an outcome target, but it should not certify that an external outcome occurred simply because the planned work was delivered.

Accounting example: an assignable item

A07 — Clarify outsourced accounting scope. Prospects repeatedly ask whether monthly close support includes an audit opinion. The approved scope excludes audit work, but the existing page does not explain the distinction. The marketing lead will revise the page, with the responsible accounting partner approving the scope and the web owner publishing it.

The brief estimates six to ten writer hours, two to three partner hours, and one to two web hours. Acceptance requires a clear description of included work, the audit exclusion, supported statements about the team, working links, and confirmation that the published page matches the approved version.

The expected client benefit is a more accurate understanding of the engagement. The hypothesized AI effect is fewer incorrect descriptions in the relevant case group. Schedule that observation separately. If a proposed case example is awaiting client permission, keep it as a dependent item; the verified scope clarification can proceed without it.

Record who maintains the page when the engagement changes. Delivery creates an ongoing information responsibility. Later chapters develop the detailed maintenance routines; the owner should be identifiable from the first work item.

7.6 Sequence Work According to Dependencies, Available Expertise, and Implementation Capacity

Commit the next delivery cycle against actual availability. Deduct routine monitoring, existing obligations, and client commitments before estimating capacity for new work. Count professional review separately from writing, design, and web implementation.

Begin with dependencies. Approved service facts may need to precede the service-page revision; permission may need to precede the case narrative; the replacement asset must be ready before its redirects are activated. Give every blocking dependency an owner and a next decision date.

Independent work can proceed concurrently. A biography correction need not wait for a research project. An owned-page correction need not wait for an external directory. Reserve reviewer time before promising publication, and keep the number of active items small enough to finish.

A worked capacity example

A consulting practice has 20 marketing hours, five practice-leader hours, and six web hours available for a two-week cycle after routine commitments. Review slots are booked; milestones use working days. P3 reorganizes P1’s approved scope. These hypothetical estimates illustrate capacity planning.

The first three items require 16 marketing hours, four practice-leader hours, and five web hours. That leaves four, one, and one hours respectively for contingencies. Adding P4 would exceed both marketing and practice-leader capacity. Give it a later review slot or revise the commitment explicitly.

P4 is a feasibility decision, not the full research study. If approved, the study needs its own scope and resource commitment. Its strategic value does not disappear because it is outside this cycle; keep its sponsor and next decision visible.

If a new material correction requires more review time than the reserved hour, revise the plan. P3 might move to the next cycle. Record the changed milestone and reason rather than keeping a completion date that the team can no longer meet.

Manage delivery and observation separately

Use a short weekly delivery check: what is blocked, ready for review, ready to publish, or awaiting an outside response? Resolve the next action and owner. Use the monthly monitoring review established in Chapter 6 to reassess findings and select subsequent work.

A practical status set is ready, in progress, in review, implemented, awaiting outside response, closed, and parked. Add the blocking reason where needed. Keep implementation status distinct from the status of follow-up observations, so delivered work does not remain indefinitely open while the team waits for an assistant to change.

Set follow-up dates according to the change and observed processing conditions. Google says recrawling can take days to weeks, that a request does not guarantee inclusion, and that repeated requests do not speed the process. Those statements do not provide a deadline for an AI answer to update. A review date is a commitment to inspect evidence. 

When an answer remains unchanged, check whether the work was implemented correctly, whether the relevant source has been revisited where that is observable, and whether the original explanation still fits the evidence. Choose a bounded next investigation if needed. Retain the uncertainty rather than repeatedly editing accurate material to chase a single answer.

Reprioritize when facts, client needs, dependencies, or capacity change. Keep a reason and review triggers for parked items. Close deliverables against their acceptance criteria and retain the linked monitoring record. The next chapters provide the detailed methods for improving firm descriptions, professional profiles, access, content, and evidence.

Practical Output: Build the Prioritized Implementation Backlog

Bring the program owner, the person responsible for publishing, and the necessary practitioners together. Use the existing gap register to produce one backlog and a realistic commitment for the next work cycle.

1. Group the findings. Link repeated observations to underlying work items. Record verified defects, unresolved causes, and affected assets.
2. Set priorities. Identify material corrections, apply publication requirements, and rank the remaining items with a short rationale covering relevance, value, confidence, effort, and risk.
3. Specify delivery. Give each selected item an owner, reviewer, deliverable, effort estimated by role, dependencies, milestone, and acceptance criteria.
4. State expected outcomes. Separate the client benefit and proposed AI effect from the controlled delivery commitment. Link the cases and dates for follow-up.
5. Commit capacity. Select the items that fit the available people and resolve their prerequisites. Reserve time for review and unexpected corrections; assign a next decision to important deferred work.

At the end of the meeting, each owner should know what to deliver next, what they need from others, who can approve it, and what evidence will establish completion. The firm should also be able to explain why this work takes precedence over the other opportunities in the backlog.