Grace Wright

Assistants lift a short passage and hand it to a buyer, so an answer in paragraph six is effectively invisible.

• Wright's fix is to answer the buying question in the first two sentences and write subheads as specific claims.
• They pair that with structural changes: real numbers in real HTML tables rather than PDFs, one page per buying question rather than per product line, and honest comparisons with named competitors on your own page.
• Wright acknowledges that the last change feels uncomfortable, but it decided the recommendation in the audit's best-performing case.

Wright contrasts "24-hour lead times on critical items" with "Responsive service," and "Holds tolerance in a 200°F washdown" with "Built to last." On one fabricator's site, the at-a-glance spec blocks they say were quoted back to them almost verbatim across three separate runs.

Read original · Original 2 · Original 3

A company writes five questions the way a buyer would actually ask them, drawing the phrasing from real sales-call quotes, then asks all five in ChatGPT, Claude, Perplexity and Gemini.

• Each answer is scored as named and recommended, cited only, used without credit, or absent.
• Wright frames the first run as a baseline to repeat month to month, so technical fixes and new content can be judged against changes in mentions.
• They are explicit that the short version gets roughly 80% of the insight at a fraction of the effort.

In Wright's own audit of 11 manufacturing websites, every site was invisible in AI answers for at least one core buying question it should have owned, and two were absent from every non-branded buying question they tested. The full audit took up to two hours per site; Wright says the do-it-yourself version takes about 15 minutes.

Read original · Original 2

That makes the findings usable as a work plan: the most common failures are also the ones most sites can start fixing this week, beginning with answer-ready page copy, schema markup that says what the company actually makes, and corroboration somewhere the company does not own.

Schema missing or generic, a third-party footprint too thin for an assistant to find corroboration, and pages an assistant cannot lift a clean answer from each appeared on 11 of 11 sites; stale money pages on 10 of 11; self-contradicting facts on 9 of 11 — including one manufacturer that spelled its flagship product two ways across its own domains.

• One company had about 39,000 words in PDFs against 8,000 in HTML, with every case study PDF-only.
• Another had a case-studies page reading "Case Study Test One" live and crawlable, and one still carried a COVID "we're open" notice in the page source.

Read original · Original 2

Averaging that variance, they say, produces a figure you cannot take apart or act on. Wright's alternative is a loop: run a technical audit alongside a prompt basket for a baseline, turn it into a ranked action plan, do the work, re-audit to isolate each action's incremental effect, then reinvest in what moved the numbers.

Wright cites a SparkToro test of 2,961 responses that found under a 1-in-100 chance that two runs of the same prompt return the same list of brands.

• Their bottom-line measure is more concrete: prospects arriving and saying they heard about the company from ChatGPT.
• That's why Wright recommends adding "how did you hear about us" to high-intent forms and having sales ask which question the buyer searched.

Read original · Original 2

They expect the human contribution to concentrate in a few places: discernment about whether output was worth making, accountability to outcomes rather than instructions, building quality systems, and persuading people to move. Wright says they understand why others read the same headlines and reach a more fearful conclusion; their own account of the shift starts with panic, not enthusiasm.

Wright describes building a Claude skill from notes they had written while doing a 12-hour account-growth analysis by hand.

• Running the skill matched those 12 hours in minutes and caught things they had missed.
• Their reaction, they write, "was not joy" — then they chose curiosity over putting their head in the sand.

Read original

What distinguishes a contractor is whether the architect, subs and general contractor can solve each problem without finger-pointing or ego — negotiating without burning relationships.

• The reasoning is local and commercial: the St.
• Louis construction community is small, and the same owners, architects and subcontractors work together repeatedly.
• Fairness on this job makes the next one easier to staff and run.

Wright quotes a subcontractor's message praising BEX for looking out for subcontractors' interests as well as the customer's and saying that "attitude brings job-site performance and quality." The subcontractor's words, relayed by Wright, provide the proof rather than Wright's own claim.

Read original · Original 2

They argue that people take chances on those they know they can trust, that only genuinely large problems should be escalated so people listen when they are, and that it pays to understand how the company makes money and who writes the checks. Wright also documents their own plans rather than keeping them private, offering to share the spreadsheets behind them.

Wright's account of moving into their family's construction business starts with months of reading and conversations, followed by a 90-day plan broken into three goals — skill up in construction, build meaningful relationships, add value immediately.

• Each goal was broken into specifics, such as learning to estimate a job alongside the project managers, weekly site visits, and one-to-ones with everyone in the company, all tracked in a personal spreadsheet.
• Separately, Wright recalls being told one minute before a client pitch that they were presenting — and winning the client because the preparation was already done.

Read original · Original 2

Start here for the broadest picture of Wright's audit work: a ranked list of failures with a fix attached to each one, showing newcomers whether the problems resemble their own.

Grace Wright audited roughly 1,800 pages on 11 manufacturing websites and reports that every site was invisible in AI answers for at least one core buying question it should have owned, with two sites absent from every non-branded question they tested.

• They rank the problems by how many sites had them: schema markup that was missing or too generic (11 of 11); a third-party footprint too thin to earn trust (11 of 11); pages an assistant cannot lift a clean answer from (11 of 11); stale content signals, including money pages last updated in 2017 (10 of 11); contradictory facts published on a company's own site (9 of 11); orphaned pages and broken proof, such as a homepage case study that 404s (9 of 11); the best proof locked in PDFs, including one case with 39,000 words in PDFs against 8,000 in HTML (8 of 11); crawlers blocked or throttled without anyone knowing (8 of 11); missing or JavaScript-injected H1s (3 of 11); and live placeholder content, like a case-studies page reading "Case Study Test One" (2 of 11).
• Each issue comes with a specific fix.

For brands planning a campaign

Useful for scoping an engagement around prevalence rather than guesswork. The top four issues appear on nearly every site they examined, so a fix plan can be sequenced by how often each problem occurs, starting with answer-ready page copy, schema, and third-party corroboration.

Start here to see the self-serve method behind Wright's published findings and judge whether their approach to measuring visibility matches your own.

Wright describes their method, scaled down so a company can run it alone in about 15 minutes.

• Write five questions the way a buyer would ask them, using phrasing from real sales-call quotes rather than keyword tools; ask all five in ChatGPT, Claude, Perplexity and Gemini; then score each answer as named and recommended, cited only, used without credit, or absent.
• The full audit took up to two hours per site using a custom skill, but roughly 80% of the insight is reachable in the short version.
• They frame the first run as a baseline to repeat monthly.

For brands planning a campaign

Sets up a before-and-after measure that does not depend on a vendor's score: a fixed basket of buyer questions, asked across four assistants and scored against the same four states each cycle. It also explains why Wright pulls prompt language from sales calls rather than keyword research.

Read this after the problem list for the constructive side of Wright's work, especially if the goal is to rewrite a small number of high-value pages.

Wright describes five features shared by the pages that earned citations and recommendations in Wright's audit.

• A direct answer in the first two sentences, with a specific differentiator such as a stated payback figure, so an assistant can hand the opening to a buyer as-is.
• A spec table in real HTML, with numbers a fabricator had quoted back to Wright almost verbatim across three separate runs.
• Named competitors compared honestly on the company's own page, which Wright describes as uncomfortable but the reason that page won the recommendation for questions about what separates two suppliers.
• Real customers named with real locations, in place of "leading manufacturers trust us."
• One page per buying question, rather than one page per product line, with product pages linking out to more specific answers.

For brands planning a campaign

Works as a page-level brief for a manufacturer's money pages.

• The competitor-comparison FAQ is the least conventional item and, in Wright's account, the one that decided a recommendation.
• Decide deliberately whether a client will publish that kind of page.

The most actionable item in the audit series, it also doubles as a quick diagnostic: most manufacturers can check each of the seven against their own site in an afternoon.

A fix list ordered for immediate action.

• Answer key buying questions succinctly.
• If a subhead asks a question, give the bottom-line answer straight away: assistants lift passages rather than whole pages, so an answer buried in paragraph six is effectively invisible.
• Make subheads specific: "24-hour lead times on critical items" instead of "Responsive service."
• Move spec tables out of PDFs and into HTML, since a competitor's worse spec table beats a better one trapped in a PDF.
• Confirm H1s are present in the HTML.
• Publish one version of every fact, after finding 9 of 11 sites contradicting themselves.
• Meaningfully update money pages, citing an Ahrefs study of about 16.975 million URLs that found ChatGPT cites content 458 days newer than Google organic does.
• Make key leaders visible: one company employed an ASTM standards-committee chair with no page that said so, and another had no extractable leadership, so an assistant pulled a wrong CEO name from a data broker.

For brands planning a campaign

Forms a quick-win backlog for a first engagement, especially specific-subhead rewrites, spec-table migration, and leader bio pages, which Wright treats as among the cheapest trust signals available. Useful when a client needs visible progress before approving a longer content program.

This is the measurement item in the series and the clearest statement of how Wright separates a tool's score from visible progress.

Wright argues against single-number visibility metrics because the same question returns different brands depending on whether ChatGPT is logged in, which model is used, and even a 15-minute gap between runs.

• They cite a SparkToro test of 2,961 responses that found under a 1-in-100 chance that two runs of the same prompt return the same list of brands.
• Their replacement is a cycle: run a technical SEO audit alongside a prompt basket to get a baseline; turn that into a ranked action plan, from technical fixes to new content to off-site trust signals; do the work; re-audit to see the incremental impact of each action; and feed the lessons into the next round of prioritisation.
• They argue the only real measure is prospects arriving and saying they heard about the company from ChatGPT, and recommend adding "how did you hear about us" to high-intent forms so sales can capture the exact questions buyers asked.

For brands planning a campaign

Sets reporting expectations before a project starts: use a repeatable prompt basket instead of a vendor score, and add attribution questions to forms and sales scripts so results can be traced to real inquiries. Helpful for pushing back on dashboards that cannot be unpacked by question, platform, or date.

Worth opening first if the audit series reads as purely tactical: it explains the stance behind their work and how they handle the anxiety manufacturers feel about the topic.

Wright describes how their view of AI changed.

• Returning from maternity leave after ChatGPT's breakout, they initially filed it as a "cool party trick" because checking and correcting its output took as long as doing the work.
• Later, they fed their own written notes from a 12-hour account-growth analysis into Claude, built a skill from their decision-making, and ran it.
• It matched the 12 hours of work in minutes and caught things they had missed.
• They describe the reaction as fear rather than joy, followed by a choice between putting their head in the sand and learning as much as possible; they chose curiosity.
• They list four things they think humans still do better: discernment about whether output was worth producing, accountability to outcomes, building systems centred on quality, and persuading people to go where they would not go alone.

For brands planning a campaign

Useful for message framing when a client's team is uneasy about AI search. Wright's framing is that output is becoming infinite while attention stays scarce, supporting a strategy built on judgement, accountability, and quality rather than content volume.

This is the clearest window into their operating principles and communication style, useful to prospective clients and collaborators before they read their audit findings.

Wright says they rose from intern to director of strategy at the same marketing consulting firm over eight years, and from the bottom rung to head of operations. They list the habits they wish they had known at the start: never gossip, defined as talking about someone behind their back without intending a direct conversation; always follow through or follow up; only escalate "five-alarm fires" so people listen when you do; prepare, then roll with the punches, illustrated by being told one minute before a client pitch that they were presenting—and winning the client; solicit honest feedback and act on it; ask "what am I not seeing?"; learn to negotiate, citing Never Split the Difference; and know how the company makes money and who writes the checks.

For brands planning a campaign

Relevant when judging working style: Wright emphasises preparation, directness, prioritising work that drives revenue, and tolerance for clients who are paying for a result. Sets expectations for how a project with them would likely be run.

Open this early if the two halves of their work otherwise look unrelated.

Wright describes their approach to joining BEX, a construction management firm, as eating an elephant one bite at a time.

• After months of reading and talking to people who knew more than they did, they settled on three first-90-day goals: skill up in construction by learning to estimate a job with project managers and visiting sites weekly; build meaningful relationships through one-to-ones with everyone at the company and a personal CRM for follow-ups; and add immediate value without a preconceived answer, starting with a win/loss analysis and continuing to post.
• They note that the plan lives in a spreadsheet they offer to share with others making a similar transition.

For brands planning a campaign

Explains why Wright can speak to both sides of a manufacturer's problem: marketing and search practice from their agency years, plus the client-side view of construction buyers, subcontractors, and project schedules. Relevant for campaigns that need someone credible with both audiences.

This is the clearest sample of their industry writing, distinct from their search-marketing posts, and shows the arguments they make for a construction audience.

Writing from their role at BEX, Wright argues that problems arise on every construction project, regardless of the quality of the drawings, subcontractors, general contractor, or schedule.

• The difference, they say, is whether the architect, subcontractors, and general contractor can solve each problem without finger-pointing or ego.
• Wright calls the ability to negotiate without burning relationships one of a good general contractor's most important skills.
• BEX works repeatedly with the same subcontractors, owners, and architects in the small St.
• Louis market, so the company aims to build a reputation for fairness that carries into the next problem.
• Wright quotes a subcontractor's message praising BEX for looking out for subcontractors' interests as well as the customer's, and linking that attitude to job-site performance and quality.

For brands planning a campaign

Shows how Wright turns a business position into a plain, example-led argument, using a customer quote as proof—the same pattern they recommend for web pages that AI assistants can lift and cite.

They report that every site examined was invisible in AI answers for at least one core buying question it should have owned, and that two were absent from every non-branded buying question tested. The series covers the ten most common problems found, a do-it-yourself visibility test, seven fixes to make first, the page structure that earns citations, and how to tell whether the work had an effect.

Wright pairs a page-level technical and content review with a 'prompt basket' of five questions written the way a buyer would ask them, run in ChatGPT, Claude, Perplexity and Gemini.

• Each answer is scored as named and recommended, cited only, used without credit, or absent.
• They say the full AI-assisted audit took up to two hours per site, while a shorter version takes about 15 minutes.

The audit samples 11 sites, so ask for broader validation before treating it as a benchmark. Wright notes that the 15-minute version yields roughly 80% of the insight.

For brands planning a campaign

A manufacturer buying an AI-visibility engagement gets a specific, repeatable scope: a baseline of buyer-style prompts run across the major assistants, a ranked fix list ordered by speed of payoff, and a re-audit cycle to show incremental impact. Because Wright publishes their method, the work can be checked against the standard they set.

Examples: Source 1 · Source 2 · Source 3 · Source 4 · Source 5

They also cover schema markup, thin third-party footprints, orphaned pages and broken proof, crawler blocking and throttling, and live placeholder content.

For brands planning a campaign

Brands can commission and measure this content-and-technical workstream: a crawl and self-audit, page-level rewrites of the highest-value pages, PDF-to-HTML migration of spec and case-study assets, and leadership bio pages. Wright provides copy-level before-and-after examples, making the campaign easy to write and the deliverable easy to inspect.

Examples: Source 1 · Source 2 · Source 3

They cite a SparkToro test of 2,961 responses that found under a 1-in-100 chance that two runs return the same list.

• Their alternative is a cycle: baseline technical and prompt audit, ranked action plan, implementation, re-audit to isolate what moved, then reinvestment in what worked.
• They also argue that real prospects saying they heard about the company from ChatGPT is the only measure that matters, and recommend adding 'how did you hear about us' to high-intent forms and having sales ask what buyers searched.

Month-over-month re-runs of the same prompt basket, scored with the same four states, paired with a full technical and content audit.

For brands planning a campaign

Brands can buy a measurement rhythm rather than a dashboard: monthly prompt re-runs against a fixed question set, tied to form and sales attribution. It offers a useful counter-position for manufacturers being pitched a proprietary AI visibility score.

Examples: Source 1 · Source 2

They describe building a Claude skill from their own documented decision-making and seeing it match 12 hours of account work in minutes, then choosing curiosity over fear. The four human differentiators are discernment, accountability to outcomes, painstakingly built quality systems, and persuasion — with the argument that output is becoming infinite while attention stays scarce.

For brands planning a campaign

Useful for brands that want the 'where humans fit' angle alongside a technical audit, such as a webinar, event slot or thought-leadership piece aimed at teams being told to adopt AI tooling.

Examples: Source 1

Louis commercial general contractor, Wright writes as a practitioner inside the industry.

• They argue that problems are inevitable regardless of drawings, subs, schedule or site coordination; the differentiator is whether the architect, subcontractors and general contractor can solve them without finger-pointing or ego.
• In their account, a good GC absorbs tight schedules, big personalities, weather and long lead times, brings the owner a solution, and preserves relationships because the local market is small and the same firms work together repeatedly.
• They illustrate this with a subcontractor's own quoted feedback on working with BEX.

For brands planning a campaign

Construction technology, building products and trade-service brands can reach a practitioner audience through this material. Wright's angle — reputation for fairness as commercial infrastructure in a repeat-player local market — gives content aimed at contractors and specifiers a usable hook.

Examples: Source 1 · Source 2 · Source 3

They break those goals into concrete actions — working with project managers and the VP of engineering to estimate a job independently, weekly site visits, a personal CRM for follow-ups, and a win/loss analysis — and say the whole plan is kept in a spreadsheet they are willing to share.

Public documentation of a structured onboarding plan, tracked in a personal spreadsheet, with weekly field time built in.

For brands planning a campaign

A ready-made narrative for brands serving family-owned or second-generation businesses, construction software, or CRM and proposal tools. The plan is specific enough to anchor a case study, workshop or template collaboration instead of a generic leadership post.

Examples: Source 1

Their experience has been the opposite: women on the BEX leadership team, clients and leaders at other firms have been generous with access and knowledge. They recall a client telling them on a job site to ask anything, 'nothing's too dumb,' and conclude there is plenty of room for strong women in the field.

For brands planning a campaign

A well-defined audience angle for brands running recruitment, inclusion or trade-workforce campaigns, and for building-products brands whose buyers include women in construction leadership. Wright's material is first-person and specific rather than slogan-led.

Examples: Source 1 · Source 2

They list the habits that carried them: no workplace gossip; follow through or follow up; only escalate five-alarm fires; prepare thoroughly and then improvise; solicit and act on feedback; never assume you know everything; learn to negotiate; know how the company makes money and who writes the cheques; and get comfortable saying 'I don't know,' 'I need help,' 'I was wrong' and 'I'm sorry.' A separate post sets out their self-described strengths — driving employee net promoter score from -25 to 100 in a year, an extreme internal locus of control, ruthless prioritisation, and communicating complex things simply — with wins including 418% growth in gross monthly revenue over eight months and speaking at the Industrial Marketing Summit.

For brands planning a campaign

This is the softer, team-and-operations side of Wright's profile: useful for leadership development, sales-enablement or employer-brand campaigns. It shows their marketing judgment comes from operating roles and revenue accountability, not only from auditing websites.

Examples: Source 1 · Source 2

Louis commercial general contractor.

• Their posts describe learning the construction business from project managers and the VP of engineering as part of a first-90-days plan, sitting on a leadership team that includes other women, and working with subcontractors, owners, and architects on repeat projects.
• They frame the role as a job at a company they are still getting to know, not a business they own.

View example

Their posts mention a speaking appearance but do not specify the kind of engagement.

View example

Vijay's argument that starting from keywords misdirects an AEO strategy lands close to Wright's insistence that prompts come from real sales-call language. Both walk through the retrieval process behind a single question rather than treating AI visibility as a ranking position.

Vijay reasons from how models retrieve and cite information, including query fan-out and citation-source studies, and gives a step-by-step method for diagnosing a category.

• Wright draws on an audit they ran across 11 manufacturer websites, returning to the buyer's decision and the specific page copy that wins it.
• Vijay draws on other people's research and observed model behaviour; Wright draws on their own inspection of sites.

Read original · Original 2

Both treat AI search as a practical operational programme rather than a marketing slogan.

Howland writes from inside a performance advertising agency serving large direct-to-consumer brands.

• Their material is oriented to ecommerce product pages, feeds and catalogues, with advertising data from companies like Fairing as evidence.
• Wright's audit covers B2B manufacturers and their spec tables, PDFs, leadership pages and site-wide contradictions; their measurement advice is to repeat a fixed prompt set over months.

Read original · Original 2

A reader following Wright for practical website guidance would find the subject matter adjacent.

Shukla's posts are general ten-point summaries that assert good practice without showing the sites behind the advice.

• They include no audits, figures or named examples, and cover ecommerce store design rather than AI assistants and B2B manufacturing.
• Wright's equivalent material is grounded in a specific audit, with before-and-after copy such as a lead-time claim replacing 'responsive service'.

Read original · Original 2

Their approach starts with the retrieval process, not a keyword list: pick five to ten real buyer questions, run them in ChatGPT, see whether a web search happens and what it pulls, trace the fan-out searches behind one prompt (they cite an average of 7.6 per prompt), then study which sources are repeatedly cited. Their argument that keyword-first planning misdirects an AEO program—a page can rank for a seed term while missing the information an assistant assembles—suits a brand that has already invested in SEO content and wants a diagnosis before commissioning more.

The tradeoff is page-level depth.

• These frameworks and tests draw on consultation calls, not a documented site audit.
• They lack the equivalent of Wright's roughly 1,800 pages across 11 manufacturing websites, a ranked list of recurring manufacturing defects, and industrial proof such as spec tables moved into HTML or named-customer case studies.
• A manufacturer that needs concrete fixes and an industrial reference set gets more from Wright's audit series; a brand that needs to know whether AI retrieval is winnable at all gets a sharper starting point from Vijay.
• The posts don't describe audience data or a published commercial offering, so confirm reach and terms directly.

Examples: Source 1 · Source 2

The AEO playbook covers mapping 20 to 30 buying decisions from support tickets, reviews and post-purchase surveys; checking which OpenAI crawlers are allowed; keeping price, availability and returns in visible HTML that matches the merchant feed; publishing return, support and sizing data only the brand holds; writing honest comparison pages with dated methodology; and measuring through Search Console, AI referrals and an AI option in the post-purchase survey.

• The companion post reports Fairing data across 158 direct-to-consumer brands over 104 weeks: surveys caught 7.5 times more AI-influenced orders than click-based attribution and 10.6 times the revenue.
• A brand that must justify AI-search work to a finance team, or sells direct to consumers, gets the stronger case here.

The tradeoff is fit with a manufacturing buyer.

• Howland focuses on DTC ecommerce and paid-ads agency material—catalog feeds, review velocity, podcast-style attribution benchmarks—rather than industrial buying questions or spec documentation.
• Much of the measurement relies on post-purchase surveys that assume a direct consumer transaction, which many manufacturers cannot run at useful volume.
• Howland also states plainly that the playbook credits another practitioner and that they do not consider themselves an AEO expert, making the AI-search substance secondhand.
• The posts don't describe audience data or commercial terms.

Examples: Source 1 · Source 2

The two posts set out ten points each: SEO-friendly structure, mobile-first design, page speed, readable URLs, navigation, clean product pages, headings, internal links, image alt text and trust signals, plus checkout topics such as guest checkout, payment choices, transparent pricing and helpful error messages. The structure works as an outline for an introductory guide, slide deck or list of items to hand a junior web team.

The tradeoff is originality and specificity.

• The posts are generic listicles with no firsthand audit, figures from a named study or method for testing AI answers.
• They don't cover how assistants retrieve or cite content or address manufacturing buying questions.
• A manufacturer gets standard hygiene advice it could have written itself, not the diagnosis and proof a recommendation needs.
• For a manufacturer seeking diagnosis and proof, these posts are a weak substitute—best treated as low-cost introductory material rather than a basis for a decision.
• The posts don't describe audience data or published commercial terms.

Examples: Source 1 · Source 2