Before we run the audit, we need to make sure we're asking the right questions about the right competitors to the right buyers. This document presents what we've learned about Overline's market — your job is to tell us what we got right, what we got wrong, and what we missed.
Before we measure how often Overline gets named when a founder asks an assistant which seed funds write the first institutional check in Atlanta and the Southeast, these three signals tell us whether AI crawlers can reach the site, read it, and trust that it is current. All three are derived directly from the Layer 1 crawl of the four live URLs — no interpretation.
talent.overline.vc job board still has per-portfolio-company URLs sitting in search indexes, and every one of them — /companies/sola, /companies/uptimehealth, /companies, /jobs — now returns HTTP 307 to the single page overline.vc/jobs. That is the classic soft-404 pattern, and 307 is explicitly temporary, so the old URLs never consolidate.not_mentioned. But that is because robots.txt returns 404, and so does every conventional sitemap path on both hosts. With no sitemap and no robots directive, discovery rests entirely on an internal link graph of four live URLs behind a header nav of five links, four of which are same-page anchors.A founder deciding whose money to take has always built the list from three places: other founders, the local ecosystem, and a search engine. The third is being replaced — increasingly the founder types the question in plain language into an assistant and gets back four or five firm names, assembled from whatever the model can read and trust about each one. For seed venture capital that shift lands harder than it does for most categories, because the deliverable is a short list of names and the entire business is being on it: a fund that is not named in the first answer never sees the deal, and never learns it didn't.
This document is the input layer for measuring that, and it has three parts we need to validate together before anything gets run. The competitive landscape defines which firms Overline gets compared against, which determines how head-to-head queries are constructed. The buyer set defines who is asking and in whose language — a founder choosing a lead investor and an LP choosing a fund search in almost opposite vocabulary, and that distinction alone reshapes the query budget. The Layer 1 technical baseline determines something more basic: whether AI systems can extract Overline's own claims from Overline's own URLs, or have to assemble the answer from third-party trackers and press coverage instead.
The validation call is where those inputs get locked, and it produces two kinds of decisions. The first is input validation: is the audit subject scoped correctly, are the right firms in the right tiers, are the right buyers represented, and are the capability ratings honest about where Overline actually wins deals? The second is engineering triage — several of the technical findings need no decision from anyone in this document and can start the moment it is read. The Pre-Call Checklist near the end aggregates both into a single page you can work from.
talent.overline.vc with a per-path 301 map, and return 410 Gone for the /companies/* pages if no per-company destination is being restored on overline.vc./news to https://overline.vc/#news; the URL is still indexed and still being returned in search results under the title "Overline - seed stage venture capital."app/sitemap.ts and app/robots.ts at build time; nothing in the validation call changes what belongs in either file, and today neither exists at any path on either host.Three things to know before you read the rest of this document.
What this is A knowledge graph of Overline's market, built from the outside in. We read the site, the regional venture listings, the competitor sets, and the conversations Southeast founders are having about raising a first institutional round, then assembled the competing firms, buyer archetypes, capabilities and frustrations that will drive the audit's query set. One structural note: seed-stage venture capital has no G2 or Capterra to mine, so persona and pain-point language was synthesised from ecosystem sources rather than review platforms — which is exactly why the confidence badges matter here more than usual. Nothing in this document is a recommendation yet. It is the input layer.
What we need from you Corrections. Every purple box in this document is a question we genuinely cannot answer from the outside, and each one names the specific downstream consequence of getting it wrong. You know your deal flow, your LPs, and who actually sits across the table from you better than any amount of research does. Marking up this document is the highest-leverage 45 minutes in the engagement.
Confidence badges Every entity carries a confidence rating. High means it came directly from Overline's own materials or an explicit third-party listing. Medium means it was assembled from adjacent evidence — regional fund roundups, stage proximity, mandate overlap. Low means we inferred it and are flagging it for you to confirm or kill. Read the medium and low badges as an invitation to argue.
How Overline is positioned to an outside reader — and to a model reading the same pages. Everything below drives how the audit refers to the firm and what it counts as a match.
→ Is the audit subject Overline the firm, or Michael Cohn the individual investor brand? The engagement seed URL was linkedin.com/in/michaelcohn — a personal profile, not a company site — so we resolved it to Overline and set the crawl target to overline.vc, because linkedin.com is not a crawlable property and would have produced a meaningless site analysis. That resolution is a judgement call, and it is the one decision in this document that can invalidate the whole run: firm-scope means queries read "which seed funds lead rounds in Atlanta" and every mention of Overline counts as a hit; person-scope means queries read "who is Michael Cohn," the entity-normalisation rules change, and the Techstars Atlanta and Cloud Sherpas history becomes in-scope surface area rather than background. Which one are we measuring?
5 personas: 4 decision-makers, 1 influencer. These determine whose vocabulary the query set is written in — a founder choosing a lead and an LP choosing a fund do not use the same words for the same thing.
Critical review area This is the section most worth your red pen. In a normal software engagement personas come from review-platform reviewer titles; here there is no review platform, so three of these five were inferred from how Southeast founders and allocators describe the decision rather than observed directly. If a persona in this set is not real, the queries written in that persona's language are wasted budget — and we will not find that out from the results, only from you.
Data sourcing note From the knowledge graph: role, department, seniority, influence level, veto power, technical level, and the provenance behind each. Synthesised by us: the role description, primary buying jobs and query focus areas — these are our reading of what each role actually does during an investor decision, and they are the fields most likely to need correction. Personas sourced from automated_scrape reflect how Overline's own site describes who it backs; those marked llm_inference were assembled from ecosystem sources (Hypepotamus, Venture Atlanta, Startup Atlanta's funding guide, Embarc Collective's Southeast Capital Landscape Report).
→ Does the typical inbound founder arrive with angels already soft-circled and no one willing to price the round, or with nothing but a team and a deck? The first writes "who will lead my seed round in Atlanta"; the second writes "who invests in pre-revenue Atlanta startups" — and the two return different firms.
→ Is the technical co-founder genuinely co-equal on the investor choice, or does the CEO decide while the CTO diligences the operating-partner and talent side? If it is the latter, Arjun reclassifies from decision-maker to evaluator and the talent and engineering-network queries move out of the decision cluster into the diligence cluster.
→ What share of Overline's pipeline is actually second-time founders with a prior exit? Repeat founders often skip search entirely and call three people — if they are a small slice of your deal flow, that query budget is better spent on first-time-founder discovery language, and if they are a large slice, the audit should weight the terms-and-speed cluster far more heavily than the discovery cluster.
→ Do you want LP and family-office search behaviour measured in this audit, or should it stay entirely on the founder side? LPs and founders search in near-opposite language, so keeping Elaine consumes roughly a fifth of the query budget on fund-diligence queries — worth it if you are raising, pure waste if you are not.
→ Is the accelerator or ecosystem lead a buyer whose language deserves its own query cluster, or purely a referral channel that never searches? If they are a channel, that budget moves to founder-facing queries; if they are genuinely how your best deals originate, the audit should test referrer-language queries like "which Southeast funds should I send my cohort to" that no founder would ever type.
Missing personas? Three roles sometimes appear in seed venture decisions in this region — do they show up in yours? (1) The co-investor or syndicate lead at another fund — the partner searching for who is leading rounds in Atlanta so they can fill out an allocation; they search in fund language, not founder language. (2) The startup attorney or fractional CFO — often the person a first-time founder asks "is this a normal term sheet," which makes them an unlisted influencer on which offer gets accepted. (3) Corporate development or a strategic venture arm at a Southeast enterprise — relevant if co-investment alongside corporates is a real part of how rounds get assembled here. Who else shows up in your deals?
5 primary + 5 secondary firms identified. Tier assignment is not cosmetic — it decides which firms get direct head-to-head testing and which are only tested for category awareness.
Why tiers matter Primary firms each get five to eight direct-comparison queries — the "Overline vs. [firm]" constructions, plus the ones a founder actually types, like "best seed funds in Atlanta that will lead a round" and "who writes the first check for Southeast startups." That means these five tier assignments commit roughly 30–40 queries to head-to-head testing; secondary firms are tested only for category co-occurrence. We are least certain about Panoramic Ventures, which landed in the primary tier from regional venture listings rather than from observed head-to-head situations — if you rarely see them across the table, moving them to secondary frees six to eight queries for a firm you actually lose deals to.
→ Three things to correct here. (1) Who's missing? Which firms have you actually lost a lead position to in the last twelve months — including national micro-funds and solo GPs who now write first checks remotely and would never appear in an Atlanta listing? (2) Is Panoramic Ventures really primary? It was tiered from regional listings, not observed deals; if it belongs in secondary, six to eight head-to-head queries move to a firm you genuinely compete with. (3) Is Techstars Atlanta a competitor at all? Michael Cohn launched that program — so it may be a referral channel that sends you deals rather than a rival for the first check, and if so it should come out of the competitive set entirely rather than being tested for co-occurrence against you.
10 buyer-level capabilities mapped — what a founder weighs when deciding whose money to take, rated outside-in. These determine which capability queries the audit tests.
An investor who will actually lead the round, price it, and set clean terms instead of waiting for someone else to go first
Someone who will back the team before there is revenue or a formally named round, not after the metrics are already obvious
Investors who have actually built and sold a company and can tell me what to do when hiring, pricing, or the first enterprise contract goes sideways
Warm intros to real buyers at Atlanta and Southeast companies who can become my first ten customers
Can they write a big enough first check to anchor a three to five million dollar seed, or do I have to assemble the rest myself?
Will this fund still be able to write a meaningful check into my Series A and Series B, or do they tap out after the seed?
Can they get me a real meeting with a Bay Area or New York fund that leads Series As, not just a name in a spreadsheet?
Real help hiring senior engineers and go-to-market leaders, not a Slack channel and a shared job board
Do they already understand my category — AI-native services, fintech, logistics — or are they learning it on my dime?
Does having this firm on my cap table make the next round easier, or will coastal investors shrug at the name?
Which three carry the differentiation queries? The audit tests all 10 capabilities, but competitive differentiation queries will emphasise 3. Four are currently rated strong:
· Round Leadership & Term Setting — linked to 3 of the 6 high-severity buyer frustrations, the most of any capability here
· Pre-Traction Conviction (Inception-Stage Investing) — linked to 2
· Southeast Customer & Executive Network — linked to 1, and the only capability that answers the coastal-flyover problem
· Founder-Operator Experience on the Cap Table — linked to 0 high-severity frustrations, though it is the most distinctive thing in Overline's own positioning
On the pain-point linkage alone the first three win. But operator experience is what the site leads with, so the ranking and the positioning disagree — which of these best represents where Overline actually wins deals?
→ Two ratings we most want challenged. Follow-On Capital & Reserve Depth is rated weak — that came from publicly reported fund sizes (a $27.4M first fund and a $13M opportunity fund) measured against a market where seed rounds routinely exceed $3M. If current, undisclosed figures let you meaningfully follow into a Series A, we are pointing the audit at a vulnerability that does not exist and should be testing that capability offensively instead. Southeast Customer & Executive Network is rated strong — but Tech Square Ventures has Engage and Fortune 500 customer access, and Atlanta Ventures has the loudest community engine in the city; is Overline's network genuinely stronger for a founder chasing their first ten customers, or just different? Separately: is Sector Domain Depth still moderate given the AI-native professional services thesis, and should Coastal Series A Investor Access and Investor Brand Signal for the Next Round be merged, since to a founder they may be one question about what happens at the next round?
11 pain points: 6 high severity, 5 medium. The buyer language below is how the audit's queries will actually be phrased — not how a fund would describe the same problem.
Southeast founders assemble angel commitments but cannot find an institutional investor willing to lead and price the round, leaving the raise stalled indefinitely.
Regional investors defer commitments until a revenue threshold the founder cannot reach without the capital they are seeking, creating a chicken-and-egg deadlock at inception stage.
Bay Area and New York funds decline to engage with startups headquartered in the Southeast, so founders lose access to the deepest pool of seed capital purely on geography.
Companies that successfully raise seed capital in the Southeast hit a wall at Series A because few regional funds have the reserves to lead a follow-on round of that size.
Drawn-out diligence processes and unreturned follow-ups consume months of founder time and runway without producing a decision.
Limited partners question whether a small, geographically concentrated seed fund can produce venture-scale returns rather than functioning as regional economic development.
Founders receive board-level guidance from investors who have never operated a company, producing generic advice that does not survive contact with the founder's actual situation.
Seed-stage companies outside major tech hubs struggle to recruit senior engineering and go-to-market leaders, and investor 'talent support' rarely translates into actual hires.
Generalist regional funds lack pattern recognition in a founder's specific category, so they cannot benchmark metrics, pricing, or hiring against comparable companies.
A regional investor's brand carries limited weight with national Series A funds, so the seed cap table provides little validation signal in the next raise.
The strongest regional deals are sourced through closed personal networks, so founders outside those circles and co-investors without local ties never see or reach them.
→ Severity, language, and what's missing. Severity: six of eleven are rated high, which is a lot — if "the fundraise itself burns the runway" and "chicken-and-egg on traction" are actually background noise rather than the thing that makes a founder start searching, they should drop to medium and cede query weight to the ones that do. Language: the LP-side entry ("can a small Southeast-only seed fund actually return three times capital") is our lowest-confidence pain point, inferred rather than observed, and it disappears entirely if the LP persona goes out of scope — is that how allocators actually put the objection to you? Missing: three frustrations we would expect in this category but did not find evidence for — governance load ("I don't want a board seat and a monthly reporting package at half a million in revenue"), seed dilution pricing ("the local term sheet takes twenty percent when the coastal one takes twelve"), and certainty of close ("they said yes and then re-traded me at the wire"). Do any of those come up in your conversations?
Nine findings across the four live URLs on overline.vc and the retired talent.overline.vc host: 0 critical, 3 high, 5 medium, and 1 item that needs manual verification. All technical — content strategy is not in scope for this document.
Actionable now — engineering There is no emergency here: nothing is blocking AI crawlers, and every one of the seven bots we tested can reach the site today. What there is, is a rebuild that left its predecessor's URLs stranded. Three items are worth starting before the validation call. (1) The talent.overline.vc redirect map — the retired job board still has per-portfolio-company URLs in search indexes, and a blanket 307 to /jobs is telling crawlers those pages are temporarily displaced rather than permanently moved. (2) The /news 404 — one 301 to the homepage anchor recovers a URL that is still indexed and still being served in search results. (3) The missing sitemap — this one is disproportionately costly here, because Overline's genuinely fresh assets (a news feed dated 4 August 2026, a jobs board refreshed twice daily) currently have no machine-readable way to advertise that recency, and AI-cited content is 25.7% fresher on average than content appearing in traditional Google organic results (Ahrefs, August 2025). One caveat on sequencing: pull the Search Console export first — it determines whether the redirect map is a handful of URLs or several hundred.
What we found: Overline previously ran a hosted job board at talent.overline.vc with a dedicated URL per portfolio company. That board has been retired and the subdomain now points at the main Next.js application. Verified live behaviour: talent.overline.vc/, /jobs, /companies, /companies/sola and /companies/uptimehealth all return HTTP 307 (temporary redirect) to the single page https://overline.vc/jobs, while talent.overline.vc/onboarding and /cookie-policy return HTTP 404. Several of these legacy URLs are still present in search engine indexes — a site: search for overline.vc surfaced talent.overline.vc/companies/uptimehealth, /companies/sola, /companies, /jobs, /talent-network, /onboarding and /cookie-policy as indexed results.
Why it matters: Two separate problems compound here. First, many distinct source URLs redirecting to one unrelated destination is the classic soft-404 pattern: search engines and AI crawlers treat a redirect to a generic page as "this content no longer exists" rather than "this content moved here", so the accumulated link equity and citation history of the per-company pages is discarded instead of being inherited. Second, HTTP 307 is an explicitly temporary status — it instructs crawlers to keep the original URL in the index and re-check it, so the old URLs never consolidate into the new ones. For a firm whose portfolio companies are the primary entity an AI model would be asked about, losing the per-company pages removes the only Overline-owned URLs that ever associated a specific portfolio company with the Overline brand.
Recommended fix: Decide the intended destination for each legacy path and implement it as HTTP 301 (permanent), not 307. Where an equivalent destination exists (talent.overline.vc/jobs → overline.vc/jobs, talent.overline.vc/talent-network → overline.vc/talent-network), map it one-to-one. Where no equivalent exists (the per-company pages under /companies/*), either restore a per-company destination on overline.vc or return a genuine 410 Gone so crawlers de-index cleanly rather than accumulating soft 404s. Do not leave a blanket 307 to /jobs in place.
What we found: https://overline.vc/news returns HTTP 404. The URL is still indexed and was returned in search results with the title "Overline - seed stage venture capital". On the rebuilt site the news content is no longer a standalone page — it now lives in a "#news" anchor section of the homepage. No redirect from /news to the homepage anchor is in place.
Why it matters: A previously indexed URL that now hard-404s loses whatever citation and ranking history it had accumulated, and any external link or model-memorised reference pointing at /news now resolves to an error page. Portfolio-in-the-news content is exactly the kind of dated, entity-rich material AI systems cite when answering questions about which companies a firm has backed and what those companies are doing, so this is the single highest-value URL on the old site to lose.
Recommended fix: Add a 301 redirect from /news to https://overline.vc/#news so the historical URL resolves and passes its authority to the live content. Audit the rest of the pre-rebuild URL set the same way — pull the indexed URL list from Google Search Console's Pages report and map every 404 to its closest live equivalent.
What we found: https://overline.vc/sitemap.xml returns HTTP 404, as do /sitemap_index.xml, /sitemap-index.xml and /sitemap. The same is true on the talent.overline.vc host. No sitemap is referenced anywhere, since robots.txt is also absent. Page discovery therefore depends entirely on the internal link graph, which on this site consists of a header nav of five links (four of which are same-page anchors) plus /jobs, /pitch and /talent-network.
Why it matters: A sitemap is the primary mechanism for telling crawlers which URLs are canonical and when each was last modified. Without one, AI and search crawlers have no lastmod signal at all — which matters disproportionately here because freshness is a dominant citation factor and Overline's genuinely fresh content (the news feed dated to 4 August 2026, the jobs feed refreshed twice daily) has no machine-readable way to advertise that recency. It also means the crawler has no independent record that /news is gone and /jobs exists, slowing the reindexing of the rebuild.
Recommended fix: Generate a sitemap at /sitemap.xml covering all four live URLs (/, /jobs, /pitch, /talent-network) with accurate lastmod values driven by the underlying content dates — the news feed date for the homepage and the job-feed refresh timestamp for /jobs. Next.js supports this natively via an app/sitemap.ts file, so it can be generated at build time rather than maintained by hand. Submit it in Google Search Console and reference it from robots.txt once that file exists.
What we found: https://overline.vc/robots.txt returns HTTP 404. The deterministic robots parser therefore reports all seven tracked crawlers — GPTBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended, Googlebot and Bytespider — as not_mentioned. No AI crawler is blocked, which is the desired outcome, but it is the result of the file being absent rather than an explicit decision.
Why it matters: The immediate access risk is nil: with no robots.txt, every crawler is implicitly allowed and Overline's content is reachable. The cost is indirect. robots.txt is the standard location to declare the sitemap, and it is the file a client's future web vendor or security tooling is most likely to add with restrictive defaults. Establishing it now with a deliberate, permissive AI-crawler policy converts an accidental permission into an intentional one and removes the risk of a later change silently blocking GPTBot or ClaudeBot.
Recommended fix: Publish a robots.txt that explicitly allows the AI crawlers Overline wants citing it (GPTBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended, Googlebot) and declares the sitemap location with a Sitemap: https://overline.vc/sitemap.xml line. Next.js can generate this from an app/robots.ts file. Record the AI-crawler policy as a deliberate decision so it survives future site changes.
What we found: Direct inspection of the raw HTML of all four live pages found no application/ld+json blocks and no structured data of any kind — 0 schema objects on /, /jobs, /pitch and /talent-network. Notably absent: Organization markup identifying Overline, its founding date, its founders and its location; Person markup for Sean O'Brien and Michael Cohn; and JobPosting markup on /jobs, which currently renders 119 live roles across 19 portfolio companies with titles, locations, employment types and salary ranges — all of which map directly onto well-defined JobPosting fields.
Why it matters: Structured data is how a page states facts unambiguously rather than leaving them to be inferred from prose. For an investment firm the highest-value assertions are entity facts an AI is directly asked to reproduce: what Overline is, who runs it, where it is based, and when it was founded. Without Organization and Person markup those facts exist only as sentences inside a single page, so a model must infer them and is more likely to conflate Overline with similarly named entities. The /jobs case is the most concrete loss — richly structured listing data is already on the page, and JobPosting markup is the difference between that data being machine-readable and being wall text.
Recommended fix: Add JSON-LD to each page: Organization on the homepage (name, url, logo, foundingDate 2020, address in Atlanta, founder entries for both managing partners, sameAs links to the LinkedIn and X profiles already linked in the footer); Person markup for each managing partner; and JobPosting markup on /jobs generated from the same feed that already renders the listings. Validate with Google's Rich Results Test after deploying.
What we found: None of the four live pages emits a rel=canonical link element. At the same time both /jobs and /jobs/ return HTTP 200 with identical content, as do /pitch and /pitch/ — the application serves both URL forms rather than redirecting one to the other. Host normalisation is handled correctly (http://overline.vc, https://www.overline.vc and http://www.overline.vc all resolve to https://overline.vc), so the exposure is limited to path-level trailing-slash duplication rather than host duplication.
Why it matters: For a four-page site the duplicate-URL risk is small in absolute terms, but the cost of the fix is near zero and the consequence of leaving it is that crawlers must guess which form is authoritative. Where a site has only four URLs to its name, having any of them split citation signal across two variants is a disproportionate loss. A canonical tag also gives Overline an explicit, machine-readable statement of the preferred URL, which matters when third-party sites link to the trailing-slash variant.
Recommended fix: Set an explicit canonical URL on every page via the Next.js metadata API (alternates.canonical), choosing the non-trailing-slash form to match the og:url value of https://overline.vc already in use. Optionally configure trailingSlash handling so one form 301-redirects to the other rather than both returning 200.
What we found: The site has four live URLs. The homepage carries essentially all of Overline's substantive content — investment thesis, the 42-company portfolio, a dated news feed, both partner bios, the extended team, 21 operating partners and the Venture Scout Program — across roughly 880 words of body text. The header navigation exposes these as same-page anchors (/#portfolio, /#news, /#thesis, /#people) rather than as distinct URLs; probing the corresponding paths confirmed /portfolio, /news, /team, /about, /thesis and /people all return 404.
Why it matters: AI systems cite URLs, not fragments. When a model attributes a claim it names the page, so a single URL can realistically carry citation weight for one topic — and here that one URL is being asked to serve as the firm's thesis page, portfolio page, news page and team page simultaneously. Each section also gets only a short share of the page's total word count, which is why every individual topic reads as thin even though the page as a whole covers a lot of ground. The Venture Scout Program, which has its own "Learn more" call to action, has no URL of its own to send anyone to.
Recommended fix: Give the highest-value sections their own indexable URLs while keeping the homepage as the summary view — at minimum /portfolio, /news, /team and /venture-scouts. Restoring /news additionally satisfies the redirect need identified in the indexed_news_url_returns_404 finding. This is an information-architecture change to content that already exists, not a request for new content: each new URL is populated by the section already written.
What we found: https://overline.vc/pitch contains no H1 element at all. Its only heading is an H2 rendering the current form question, "What is the name of your company?". Its title tag and meta description are byte-identical to the homepage's ("Overline | The best founders in the Southeast get their first check from Overline" and the corresponding description), so the page has no distinguishing metadata. Total body text is 32 words, as the page is a multi-step form that reveals one question at a time. /talent-network is better — it has a correct H1, a unique title and a unique meta description — but carries only 94 words of body text.
Why it matters: The pitch submission page is the single most important conversion destination on the site: it is where a founder in the Southeast actually reaches Overline, and the homepage explicitly promises "we review and respond to every pitch... No black holes." As currently built it tells a crawler nothing — no H1 to establish what the page is, and metadata that makes it look like a duplicate of the homepage. A model asked how to pitch Overline has no page it can confidently cite for the answer, because the page that should hold that answer contains 32 words and identifies itself as the homepage.
Recommended fix: Add a descriptive H1 to /pitch (for example "Share your pitch with Overline") and give it a unique title and meta description. Add server-rendered introductory copy above the form covering what Overline invests in, the check size range, what happens after submission and the response commitment — content that already exists in fragments on the homepage and would make the page independently citable. Apply the same treatment to /talent-network.
The following item could not be assessed through our analysis method. We recommend your engineering team verify it manually before the validation call.
What to check: This analysis established the live HTTP behaviour of the legacy talent.overline.vc paths and of /news by fetching them directly, and confirmed several of them still appear in public search results. What it cannot establish from outside is the current index state as Google records it: how many legacy URLs remain indexed, when each was last crawled, and whether the 307 redirects are already being reported as soft 404s. Overline's own Search Console property is the only source for that.
Recommended action: In Google Search Console, open the Pages report for both the overline.vc and talent.overline.vc properties and export the full list of indexed and excluded URLs, filtering for "Not found (404)", "Soft 404" and "Page with redirect". Use that export as the authoritative input to the redirect map called for in the legacy redirect and /news findings. If talent.overline.vc is not currently verified as a property, add it.
Read the averages carefully Every score above is computed over four pages, and the freshness figure over just two of them — the other two carry no detectable date at all. That is not a sampling limitation in the usual sense: the crawl covered 100% of the live URLs, because there are only four. It does mean a single page moves any average here by a quarter, so treat these as a description of a very small surface rather than a statistical position. The two scores worth acting on regardless of denominator are schema coverage at 0.00 (a count, not an average — there is no structured data anywhere on the site) and passage extractability at 0.35, which reflects the fragment-anchor architecture rather than any individual page's writing.
Why now · AI crawler activity is compounding fast — Akamai measured 1.6 billion daily AI bot requests across its network, up 78% over six months (Akamai, February 2026). · Early citations compound: domains that AI platforms learn to trust get named more often as the answer set stabilises, and late movers inherit a list that has already formed around someone else. · What predicts citation is not what predicted rankings — brand web mentions correlate with AI citation at r = 0.664 versus 0.218 for backlinks (Seer Interactive, October 2025), which favours a firm whose name already circulates through regional press and founder conversation. · Seed venture capital in the Southeast is still early-innings in GEO: none of the firms in your competitive set is running a structured citation strategy yet, so acting now means competing against inaction rather than against an entrenched programme.
The full audit measures how often Overline gets named across the queries that actually decide this category — "who leads seed rounds in Atlanta," "which funds write the first institutional check for Southeast startups," "who invests in pre-revenue Atlanta companies," "Overline vs. other Atlanta seed funds," and the LP-side questions if that persona stays in scope — run across the selected AI platforms rather than inferred from one. What you will learn that you do not know today is specific: which of those questions return an answer naming other Atlanta firms but not Overline, what the assistant said instead, and which source it pulled from to say it. The Layer 1 fixes feed directly into that — every one landed before the audit runs raises the baseline we measure rather than the one we have to explain afterwards.
45–60 minutes walking through this document together. We work from the Pre-Call Checklist, resolve the open questions, and lock the inputs that drive the query set.
We build the buyer query set from the validated personas, competitors, capabilities and pain points, then run it across the selected AI platforms and capture every response and citation.
Visibility analysis, competitive positioning against the validated set, and a three-layer action plan — technical, content and authority — prioritised by what actually costs you citations.
Start now — engineering Three technical fixes don't depend on anything in this document being validated. (1) Build the legacy redirect map: replace the blanket 307 on talent.overline.vc with per-path 301s where a real destination exists, and return 410 Gone for the /companies/* pages if no per-company destination is being restored — and add the 301 from /news to https://overline.vc/#news in the same pass. Pull the Search Console Pages export for both properties first; it is what tells you whether this is seven URLs or several hundred. (2) Generate /sitemap.xml from app/sitemap.ts with real lastmod values off the news-feed date and the job-feed refresh timestamp, and publish app/robots.ts declaring the sitemap and explicitly allowing GPTBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended and Googlebot — no crawler is blocked today, but that is currently an accident rather than a decision. (3) Set rel=canonical on all four pages via the Next.js metadata API in the non-trailing-slash form, matching the og:url already in use. These don't depend on the rest of the audit and will improve your baseline visibility before we even measure it.
Two jobs before we meet. The questions on the left require your judgment — no one knows your business better than you. The engineering tasks on the right don't require the call at all.